{"models":[{"id":"arcee/zai-org/glm-5.3","type":"llm","name":"GLM-5.3","model_id":"zai-org/glm-5.3","capabilities":["text","tools","structured_output","thinking","reasoning"],"speed":"medium","cost":0.229,"quality":"great","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["minimal","low","medium","high","xhigh","max"],"default":"max","wire":"reasoning_effort"},"caller_params":["logit_bias","repetition_penalty","top_a"]},"pricing":{"billing_unit":"per_mtok","input":[1.4],"output":[4.4],"cached_input":[0.26]},"region":"US","model":"glm-5.3","family":"glm-5.3","pooled":true,"provider":{"slug":"arcee","display_name":"Arcee","owner_country":"US"},"service_route":{"id":"arcee/api-us","adapter":"arcee"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"United States","compliance":{"residency":"US","country":"United States","inference_location":"US","zdr":{"logging":true,"moderation":true,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"abuse_monitoring","moderation":"default_on","caching":"unknown","content_storage":"retained","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"arcee/api-us"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"arcee/zai-org/glm-5.3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"zai-org/glm-5.3-flash","capabilities":["text","tools","thinking","vision","reasoning","structured_output"],"speed":"fast","cost":0.047,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["minimal","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["logit_bias","repetition_penalty","top_a","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.15],"output":[0.5],"cached_input":[0.03]},"region":"US","model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"provider":{"slug":"arcee","display_name":"Arcee","owner_country":"US"},"service_route":{"id":"arcee/api-us","adapter":"arcee"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"United States","compliance":{"residency":"US","country":"United States","inference_location":"US","zdr":{"logging":true,"moderation":true,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"abuse_monitoring","moderation":"default_on","caching":"unknown","content_storage":"retained","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"arcee/api-us"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","file","video"],"output_modalities":["text"]},{"id":"baidu:cn/glm-5.3","type":"llm","name":"GLM-5.3","model_id":"glm-5.3","capabilities":["text"],"speed":"medium","cost":0.229,"quality":"great","context_window":0,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1}},"pricing":{"billing_unit":"per_mtok","input":[1.4],"output":[4.4],"cached_input":[0.26]},"region":"CN","model":"glm-5.3","family":"glm-5.3","provider":{"slug":"baidu","display_name":"Baidu Qianfan","owner_country":"CN"},"service_route":{"id":"baidu:cn","adapter":"baidu:cn"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"China","compliance":{"residency":"GLOBAL","country":"China","inference_location":"CN","zdr":{"logging":true,"moderation":true,"caching":null,"training":null,"subprocessors":null},"training":"unknown","logging":"other","moderation":"mandatory","caching":"unknown","content_storage":"retained","dpa_available":false,"third_party_access":true,"transfer_mechanism":"unknown","route_id":"baidu:cn"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"berget/zai-org/GLM-5.3-Flash","type":"llm","name":"GLM-5.3-Flash","model_id":"zai-org/GLM-5.3-Flash","capabilities":["text","tools","thinking","vision","reasoning","structured_output"],"speed":"fast","cost":0.059,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["frequency_penalty","min_p","presence_penalty","repetition_penalty","seed","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.280375],"output":[0.56075]},"region":"EU","model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"provider":{"slug":"berget","display_name":"Berget","owner_country":"SE"},"service_route":{"id":"berget/inference-eu","adapter":"berget"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"Sweden","compliance":{"residency":"EU","country":"Sweden","inference_location":"SE","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":false},"training":"no","logging":"none","moderation":"none","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":null,"transfer_mechanism":"eu_resident","route_id":"berget/inference-eu"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp8","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","video"],"output_modalities":["text"]},{"id":"bytedance:ap/glm-5.3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"glm-5-3-flash-260828","capabilities":["text","tools","thinking","vision","reasoning","structured_output"],"speed":"fast","cost":0.047,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3-Flash on BytePlus ModelArk AP (Johor, ap-southeast-1)","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["low","high","max"],"default":"","wire":"reasoning_effort"},"caller_params":["frequency_penalty","presence_penalty","repetition_penalty"]},"pricing":{"billing_unit":"per_mtok","input":[0.15],"output":[0.5],"cached_input":[0.03]},"region":"GLOBAL","model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"provider":{"slug":"byteplus","display_name":"BytePlus","owner_country":"CN"},"service_route":{"id":"bytedance:ap","adapter":"bytedance:ap"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","storage_location":"MY","zdr":{"logging":true,"moderation":true,"caching":true,"training":false,"subprocessors":null},"training":"no","logging":"abuse_monitoring","retention_days":180,"moderation":"default_on","caching":"implicit","content_storage":"retained","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"bytedance:ap"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","file","video"],"output_modalities":["text"]},{"id":"deepinfra/zai-org/GLM-5.3","type":"llm","name":"GLM-5.3","model_id":"zai-org/GLM-5.3","capabilities":["text","tools","structured_output","thinking","reasoning"],"speed":"fast","cost":0.209,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"max","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","repetition_penalty","seed","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.9],"output":[4],"cached_input":[0.2]},"region":"US","model":"glm-5.3","family":"glm-5.3","pooled":true,"provider":{"slug":"deepinfra","display_name":"DeepInfra","owner_country":"US"},"service_route":{"id":"deepinfra/us","adapter":"deepinfra"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"United States","compliance":{"residency":"US","country":"United States","inference_location":"US","zdr":{"logging":false,"moderation":false,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"none","caching":"unknown","content_storage":"unknown","dpa_available":true,"third_party_access":null,"transfer_mechanism":"sccs","route_id":"deepinfra/us"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp4","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"deepinfra/zai-org/GLM-5.3-Flash","type":"llm","name":"GLM-5.3-Flash","model_id":"zai-org/GLM-5.3-Flash","capabilities":["text","tools","vision","thinking","reasoning","structured_output"],"speed":"fast","cost":0.047,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","repetition_penalty","seed","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.15],"output":[0.5],"cached_input":[0.03]},"region":"US","model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"provider":{"slug":"deepinfra","display_name":"DeepInfra","owner_country":"US"},"service_route":{"id":"deepinfra/us","adapter":"deepinfra"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"United States","compliance":{"residency":"US","country":"United States","inference_location":"US","zdr":{"logging":false,"moderation":false,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"none","caching":"unknown","content_storage":"unknown","dpa_available":true,"third_party_access":null,"transfer_mechanism":"sccs","route_id":"deepinfra/us"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp4","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","file","video"],"output_modalities":["text"]},{"id":"deepinfra:flex/zai-org/GLM-5.3","type":"llm","name":"GLM-5.3","model_id":"zai-org/GLM-5.3","capabilities":["text","tools","structured_output","thinking","reasoning"],"speed":"fast","cost":0.183,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"max","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","repetition_penalty","seed","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.72],"output":[3.2],"cached_input":[0.16]},"region":"US","model":"glm-5.3","family":"glm-5.3","provider":{"slug":"deepinfra","display_name":"DeepInfra","owner_country":"US"},"service_route":{"id":"deepinfra/us","adapter":"deepinfra"},"service_tier":"flex","maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"United States","compliance":{"residency":"US","country":"United States","inference_location":"US","zdr":{"logging":false,"moderation":false,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"none","caching":"unknown","content_storage":"unknown","dpa_available":true,"third_party_access":null,"transfer_mechanism":"sccs","route_id":"deepinfra/us"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp4","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"digitalocean/glm-5.3","type":"llm","name":"GLM-5.3","model_id":"glm-5.3","capabilities":["text","tools","structured_output","reasoning","thinking"],"speed":"medium","cost":0.229,"quality":"great","context_window":1048576,"max_output_tokens":1048576,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["frequency_penalty","presence_penalty","repetition_penalty","seed","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[1.4],"output":[4.4],"cached_input":[0.26]},"region":"US","model":"glm-5.3","family":"glm-5.3","pooled":true,"access":"open","provider":{"slug":"digitalocean","display_name":"DigitalOcean","owner_country":"US"},"service_route":{"id":"digitalocean/serverless-us","adapter":"digitalocean"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"United States","compliance":{"residency":"US","country":"United States","inference_location":"US","storage_location":"US","zdr":{"logging":false,"moderation":true,"caching":true,"training":false,"subprocessors":false},"training":"no","logging":"none","moderation":"mandatory","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"dpf_sccs","route_id":"digitalocean/serverless-us"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"digitalocean/glm-5.3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"glm-5.3-flash","capabilities":["text","tools","structured_output","reasoning","thinking","vision"],"speed":"fast","cost":0.047,"quality":"good","context_window":1048576,"max_output_tokens":1048576,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["frequency_penalty","presence_penalty","repetition_penalty","seed","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.15],"output":[0.5],"cached_input":[0.03]},"region":"US","model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"access":"open","provider":{"slug":"digitalocean","display_name":"DigitalOcean","owner_country":"US"},"service_route":{"id":"digitalocean/serverless-us","adapter":"digitalocean"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"United States","compliance":{"residency":"US","country":"United States","inference_location":"US","storage_location":"US","zdr":{"logging":false,"moderation":true,"caching":true,"training":false,"subprocessors":false},"training":"no","logging":"none","moderation":"mandatory","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"dpf_sccs","route_id":"digitalocean/serverless-us"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image"],"output_modalities":["text"]},{"id":"evroc/zai-org/GLM-5.3","type":"llm","name":"GLM-5.3","model_id":"zai-org/GLM-5.3","capabilities":["text","tools","structured_output","thinking","reasoning"],"speed":"medium","cost":0.254,"quality":"great","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["minimal","low","medium","high","xhigh","max"],"default":"max","wire":"enable_thinking"},"caller_params":["chat_template_kwargs","frequency_penalty","min_p","repetition_penalty","seed","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[1.401875],"output":[5.6075]},"region":"EU","model":"glm-5.3","family":"glm-5.3","provider":{"slug":"evroc","display_name":"Evroc","owner_country":"SE"},"service_route":{"id":"evroc/think-eu","adapter":"evroc"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"Sweden","compliance":{"residency":"EU","country":"Sweden","inference_location":"SE","zdr":{"logging":false,"moderation":false,"caching":null,"training":false,"subprocessors":false},"training":"no","logging":"none","moderation":"none","caching":"unknown","content_storage":"unknown","dpa_available":true,"third_party_access":null,"transfer_mechanism":"eu_resident","route_id":"evroc/think-eu"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"fireworks/glm-5p3","type":"llm","name":"GLM-5.3","model_id":"accounts/fireworks/models/glm-5p3","capabilities":["text","tools","structured_output","thinking","reasoning"],"speed":"medium","cost":0.229,"quality":"great","context_window":1040000,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["logit_bias","min_p","mirostat_lr","mirostat_target","repetition_penalty","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[1.4],"output":[4.4],"cached_input":[0.26]},"region":"GLOBAL","aliases":["fireworks/glm-5.3"],"model":"glm-5.3","family":"glm-5.3","pooled":true,"provider":{"slug":"fireworks","display_name":"Fireworks","owner_country":"US"},"service_route":{"id":"fireworks/serverless-global","adapter":"fireworks"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","zdr":{"logging":false,"moderation":false,"caching":null,"training":false,"subprocessors":true},"training":"no","logging":"none","moderation":"default_on","caching":"unknown","content_storage":"unknown","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"fireworks/serverless-global"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"fireworks/glm-5p3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"accounts/fireworks/models/glm-5p3-flash","capabilities":["text","tools","thinking","vision","reasoning","structured_output"],"speed":"fast","cost":0.047,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["logit_bias","min_p","mirostat_lr","mirostat_target","repetition_penalty","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.15],"output":[0.5],"cached_input":[0.03]},"region":"GLOBAL","model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"provider":{"slug":"fireworks","display_name":"Fireworks","owner_country":"US"},"service_route":{"id":"fireworks/serverless-global","adapter":"fireworks"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","zdr":{"logging":false,"moderation":false,"caching":null,"training":false,"subprocessors":true},"training":"no","logging":"none","moderation":"default_on","caching":"unknown","content_storage":"unknown","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"fireworks/serverless-global"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","video"],"output_modalities":["text"]},{"id":"fireworks:priority/glm-5p3","type":"llm","name":"GLM-5.3","model_id":"accounts/fireworks/routers/glm-5p3-fast","capabilities":["text","tools","structured_output","thinking","reasoning"],"speed":"medium","cost":0.283,"quality":"great","context_window":1040000,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["logit_bias","min_p","mirostat_lr","mirostat_target","repetition_penalty","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[2.1],"output":[6.6],"cached_input":[0.39]},"region":"GLOBAL","model":"glm-5.3","family":"glm-5.3","provider":{"slug":"fireworks","display_name":"Fireworks","owner_country":"US"},"service_route":{"id":"fireworks/serverless-global","adapter":"fireworks"},"service_tier":"priority","maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","zdr":{"logging":false,"moderation":false,"caching":null,"training":false,"subprocessors":true},"training":"no","logging":"none","moderation":"default_on","caching":"unknown","content_storage":"unknown","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"fireworks/serverless-global"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"greenference/glm-5.3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"greenference/glm-5.3-flash","capabilities":["text","tools","thinking","vision","reasoning","structured_output"],"speed":"fast","cost":0.046,"quality":"good","context_window":1048576,"max_output_tokens":943718,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"caller_params":["seed","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.14],"output":[0.49],"cached_input":[0.028]},"region":"EU","model":"glm-5.3-flash","family":"glm-5.3-flash","provider":{"slug":"greenference","display_name":"Greenference","owner_country":"FR"},"service_route":{"id":"greenference","adapter":"greenference"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"EU","inference_location":"EU","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":true},"training":"no","logging":"none","moderation":"none","caching":"implicit","cache_scope":"shared","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"eu_resident","route_id":"greenference"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp8","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image"],"output_modalities":["text"]},{"id":"greenpt/glm-5.3","type":"llm","name":"GLM-5.3","model_id":"glm-5.3","capabilities":["text","structured_output","tools","reasoning"],"speed":"medium","cost":0.237,"quality":"great","context_window":1000000,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high"],"default":"","wire":"reasoning_effort"},"caller_params":["min_p","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[1.2336500000000001],"output":[4.9346000000000005],"cached_input":[0.30841250000000003]},"region":"EU","model":"glm-5.3","family":"glm-5.3","pooled":true,"provider":{"slug":"greenpt","display_name":"GreenPT","owner_country":"NL"},"service_route":{"id":"greenpt/eu","adapter":"greenpt"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"EU","inference_location":"EU","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":true},"training":"no","logging":"none","moderation":"unknown","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"eu_resident","route_id":"greenpt/eu"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"greenpt/glm-5.3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"glm-5.3-flash","capabilities":["text","structured_output","tools","vision","reasoning"],"speed":"fast","cost":0.045,"quality":"good","context_window":1000000,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high"],"default":"","wire":"reasoning_effort"},"caller_params":["min_p","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.12336499999999999],"output":[0.49345999999999995],"cached_input":[0.024672999999999997]},"region":"EU","model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"provider":{"slug":"greenpt","display_name":"GreenPT","owner_country":"NL"},"service_route":{"id":"greenpt/eu","adapter":"greenpt"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"EU","inference_location":"EU","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":true},"training":"no","logging":"none","moderation":"unknown","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"eu_resident","route_id":"greenpt/eu"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","video"],"output_modalities":["text"]},{"id":"inceptron/zai-org/GLM-5.3","type":"llm","name":"GLM-5.3","model_id":"zai-org/GLM-5.3","capabilities":["text","tools","structured_output","thinking","reasoning"],"speed":"fast","cost":0.183,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","low","high","max"],"default":"","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","presence_penalty","repetition_penalty","seed","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.55],"output":[3.39],"cached_input":[0.165]},"region":"EU","model":"glm-5.3","family":"glm-5.3","pooled":true,"provider":{"slug":"inceptron","display_name":"Inceptron","owner_country":"SE"},"service_route":{"id":"inceptron/inference-eu","adapter":"inceptron"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"EU","inference_location":"EU","zdr":{"logging":false,"moderation":false,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"default_on","caching":"unknown","content_storage":"unknown","dpa_available":true,"third_party_access":true,"transfer_mechanism":"eu_resident","route_id":"inceptron/inference-eu"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp4","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"inceptron/zai-org/GLM-5.3-Flash","type":"llm","name":"GLM-5.3-Flash","model_id":"zai-org/GLM-5.3-Flash","capabilities":["text","tools","thinking","vision","reasoning","structured_output"],"speed":"fast","cost":0.045,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","low","high","max"],"default":"","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","presence_penalty","repetition_penalty","seed","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.16],"output":[0.45],"cached_input":[0.07]},"region":"EU","model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"provider":{"slug":"inceptron","display_name":"Inceptron","owner_country":"SE"},"service_route":{"id":"inceptron/inference-eu","adapter":"inceptron"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"EU","inference_location":"EU","zdr":{"logging":false,"moderation":false,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"default_on","caching":"unknown","content_storage":"unknown","dpa_available":true,"third_party_access":true,"transfer_mechanism":"eu_resident","route_id":"inceptron/inference-eu"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp8","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","video"],"output_modalities":["text"]},{"id":"lyceum/z-ai/glm-5.3","type":"llm","name":"GLM-5.3","model_id":"z-ai/glm-5.3","capabilities":["text","tools","structured_output","reasoning","thinking"],"speed":"medium","cost":0.229,"quality":"great","context_window":1048576,"max_output_tokens":65536,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["minimal","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"}},"pricing":{"billing_unit":"per_mtok","input":[1.4],"output":[4.4],"cached_input":[0.26]},"region":"EU","model":"glm-5.3","family":"glm-5.3","provider":{"slug":"lyceum","display_name":"Lyceum","owner_country":"DE"},"service_route":{"id":"lyceum","adapter":"lyceum"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"EU","inference_location":"EU","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"none","caching":"implicit","cache_scope":"organization","content_storage":"ephemeral","dpa_available":true,"third_party_access":null,"transfer_mechanism":"eu_resident","route_id":"lyceum"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"lyceum/z-ai/glm-5.3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"z-ai/glm-5.3-flash","capabilities":["text","tools","structured_output","reasoning","thinking","vision"],"speed":"fast","cost":0.05,"quality":"good","context_window":1048576,"max_output_tokens":65536,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["minimal","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"}},"pricing":{"billing_unit":"per_mtok","input":[0.2],"output":[0.5],"cached_input":[0.05]},"region":"EU","model":"glm-5.3-flash","family":"glm-5.3-flash","provider":{"slug":"lyceum","display_name":"Lyceum","owner_country":"DE"},"service_route":{"id":"lyceum","adapter":"lyceum"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"EU","inference_location":"EU","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"none","caching":"implicit","cache_scope":"organization","content_storage":"ephemeral","dpa_available":true,"third_party_access":null,"transfer_mechanism":"eu_resident","route_id":"lyceum"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image"],"output_modalities":["text"]},{"id":"melious/glm-5.3","type":"llm","name":"GLM-5.3","model_id":"glm-5.3","capabilities":["text","tools","thinking","reasoning","structured_output"],"speed":"medium","cost":0.198,"quality":"great","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"max","wire":"reasoning_effort"},"caller_params":["frequency_penalty","presence_penalty","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[1.1215],"output":[3.3644999999999996],"cached_input":[0.2243]},"region":"EU","model":"glm-5.3","family":"glm-5.3","pooled":true,"provider":{"slug":"melious","display_name":"Melious","owner_country":"DE"},"service_route":{"id":"melious/eu","adapter":"melious"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"Finland","compliance":{"residency":"EU","country":"Finland","inference_location":"FI","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":true},"training":"no","logging":"none","moderation":"none","caching":"implicit","cache_scope":"shared","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"eu_resident","route_id":"melious/eu"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"melious/glm-5.3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"glm-5.3-flash","capabilities":["text","tools","thinking","vision","reasoning","structured_output"],"speed":"fast","cost":0.042,"quality":"good","context_window":1000000,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["low","high","max"],"default":"","wire":"reasoning_effort"},"caller_params":["top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.11215],"output":[0.4486],"cached_input":[0.02243]},"region":"EU","model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"provider":{"slug":"melious","display_name":"Melious","owner_country":"DE"},"service_route":{"id":"melious/eu","adapter":"melious"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"Finland","compliance":{"residency":"EU","country":"Finland","inference_location":"FI","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":true},"training":"no","logging":"none","moderation":"none","caching":"implicit","cache_scope":"shared","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"eu_resident","route_id":"melious/eu"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","video"],"output_modalities":["text"]},{"id":"mistral/zai-glm-5-3","type":"llm","name":"GLM-5.3","model_id":"zai-glm-5-3","capabilities":["text","tools","structured_output","thinking"],"speed":"medium","cost":0.241,"quality":"great","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"caller_params":["frequency_penalty","presence_penalty","repetition_penalty","seed","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[1.54],"output":[4.84],"cached_input":[0.154]},"region":"EU","aliases":["mistral:eu/zai-glm-5-3"],"model":"glm-5.3","family":"glm-5.3","pooled":true,"provider":{"slug":"mistral","display_name":"Mistral","owner_country":"FR"},"service_route":{"id":"mistral/api-eu","adapter":"mistral:eu"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"EU","inference_location":"EU","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"none","caching":"explicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":null,"transfer_mechanism":"eu_resident","route_id":"mistral/api-eu"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"mistral:global/zai-glm-5-3","type":"llm","name":"GLM-5.3","model_id":"zai-glm-5-3","capabilities":["text","tools","structured_output","thinking"],"speed":"medium","cost":0.229,"quality":"great","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"caller_params":["frequency_penalty","presence_penalty","repetition_penalty","seed","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[1.4],"output":[4.4],"cached_input":[0.14]},"region":"GLOBAL","model":"glm-5.3","family":"glm-5.3","provider":{"slug":"mistral","display_name":"Mistral","owner_country":"FR"},"service_route":{"id":"mistral:global","adapter":"mistral"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"none","caching":"explicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":null,"transfer_mechanism":"sccs","route_id":"mistral:global"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"modal/glm-5.3","type":"llm","name":"GLM-5.3","model_id":"opper--ep-glm-5-3-server.us-west.modal.direct","capabilities":["text","tools","structured_output","reasoning","thinking"],"speed":"medium","cost":0.229,"quality":"great","context_window":1048576,"max_output_tokens":1048576,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"}},"pricing":{"billing_unit":"per_mtok","input":[1.4],"output":[4.4],"cached_input":[0.26]},"region":"GLOBAL","thinking":true,"model":"glm-5.3","family":"glm-5.3","access":"open","provider":{"slug":"modal","display_name":"Modal","owner_country":"US"},"service_route":{"id":"modal/shared-us-west","adapter":"modal-shared"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":null},"training":"opt_in","logging":"none","moderation":"none","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"modal/shared-us-west"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"modal/glm-5.3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"opper--ep-glm-5-3-flash-server.us-west.modal.direct","capabilities":["text","tools","structured_output","vision","reasoning","thinking"],"speed":"fast","cost":0.115,"quality":"good","context_window":1048576,"max_output_tokens":1048576,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"}},"pricing":{"billing_unit":"per_mtok","input":[0.45],"output":[1.5],"cached_input":[0.09]},"region":"GLOBAL","thinking":true,"model":"glm-5.3-flash","family":"glm-5.3-flash","access":"open","provider":{"slug":"modal","display_name":"Modal","owner_country":"US"},"service_route":{"id":"modal/shared-us-west","adapter":"modal-shared"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":null},"training":"opt_in","logging":"none","moderation":"none","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"modal/shared-us-west"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","video"],"output_modalities":["text"]},{"id":"morph/morph-glm53-744b","type":"llm","name":"GLM-5.3","model_id":"morph-glm53-744b","capabilities":["text","tools","structured_output","thinking","reasoning"],"speed":"fast","cost":0.235,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","frequency_penalty","presence_penalty","repetition_penalty","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.082],"output":[6],"cached_input":[0.051]},"region":"US","model":"glm-5.3","family":"glm-5.3","pooled":true,"provider":{"slug":"morph","display_name":"Morph","owner_country":"US"},"service_route":{"id":"morph","adapter":"morph"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"United States","compliance":{"residency":"US","country":"United States","inference_location":"US","zdr":{"logging":true,"moderation":true,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"abuse_monitoring","retention_days":30,"moderation":"default_on","caching":"unknown","content_storage":"retained","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"morph"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp8","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"morph/morph-glm53flash","type":"llm","name":"GLM-5.3-Flash","model_id":"morph-glm53flash","capabilities":["text","tools","thinking","vision","reasoning","structured_output"],"speed":"fast","cost":0.053,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","frequency_penalty","presence_penalty","repetition_penalty","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.093],"output":[0.65],"cached_input":[0.033]},"region":"US","model":"glm-5.3-flash","family":"glm-5.3-flash","provider":{"slug":"morph","display_name":"Morph","owner_country":"US"},"service_route":{"id":"morph","adapter":"morph"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"United States","compliance":{"residency":"US","country":"United States","inference_location":"US","zdr":{"logging":true,"moderation":true,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"abuse_monitoring","retention_days":30,"moderation":"default_on","caching":"unknown","content_storage":"retained","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"morph"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp8","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","video"],"output_modalities":["text"]},{"id":"nebius/zai-org/GLM-5.3","type":"llm","name":"GLM-5.3","model_id":"zai-org/GLM-5.3","capabilities":["text","tools","structured_output","thinking","reasoning"],"speed":"medium","cost":0.229,"quality":"great","context_window":1024000,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["minimal","low","medium","high","xhigh","max"],"default":"max","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs"]},"pricing":{"billing_unit":"per_mtok","input":[1.4],"output":[4.4]},"region":"EU","model":"glm-5.3","family":"glm-5.3","pooled":true,"provider":{"slug":"nebius","display_name":"Nebius","owner_country":"NL"},"service_route":{"id":"nebius/studio-eu","adapter":"nebius"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","storage_location":"FI","zdr":{"logging":true,"moderation":false,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"other","moderation":"none","caching":"unknown","content_storage":"retained","dpa_available":true,"third_party_access":null,"transfer_mechanism":"sccs","route_id":"nebius/studio-eu"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp4","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"nebius/zai-org/GLM-5.3-Flash","type":"llm","name":"GLM-5.3-Flash","model_id":"zai-org/GLM-5.3-Flash","capabilities":["text","tools","thinking","vision","reasoning","structured_output"],"speed":"fast","cost":0.047,"quality":"good","context_window":1024000,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["minimal","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs"]},"pricing":{"billing_unit":"per_mtok","input":[0.15],"output":[0.5]},"region":"EU","model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"provider":{"slug":"nebius","display_name":"Nebius","owner_country":"NL"},"service_route":{"id":"nebius/studio-eu","adapter":"nebius"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","storage_location":"FI","zdr":{"logging":true,"moderation":false,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"other","moderation":"none","caching":"unknown","content_storage":"retained","dpa_available":true,"third_party_access":null,"transfer_mechanism":"sccs","route_id":"nebius/studio-eu"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp4","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image"],"output_modalities":["text"]},{"id":"nextbit:global/glm:5.3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"glm:5.3-flash","capabilities":["text","tools","thinking","reasoning","structured_output"],"speed":"fast","cost":0.051,"quality":"good","context_window":1048576,"max_output_tokens":128000,"description":"GLM-5.3-Flash on Nextbit's global endpoint; fp8, 1M-token context, served text-in only","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","min_p","repetition_penalty","seed","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.165],"output":[0.55],"cached_input":[0.033]},"region":"GLOBAL","aliases":["nextbit/glm:5.3-flash"],"model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"provider":{"slug":"nextbit","display_name":"Nextbit","owner_country":"ES"},"service_route":{"id":"nextbit:global","adapter":"nextbit:global"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":true},"training":"no","logging":"none","moderation":"unknown","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"nextbit:global"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp8","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","file","video"],"output_modalities":["text"]},{"id":"novita/zai-org/glm-5.3","type":"llm","name":"GLM-5.3","model_id":"zai-org/glm-5.3","capabilities":["text","tools","thinking","reasoning","structured_output"],"speed":"medium","cost":0.229,"quality":"great","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"max","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","repetition_penalty","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[1.4],"output":[4.4],"cached_input":[0.26]},"region":"US","model":"glm-5.3","family":"glm-5.3","pooled":true,"provider":{"slug":"novita","display_name":"Novita","owner_country":"US"},"service_route":{"id":"novita/api-us","adapter":"novita"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"United States","compliance":{"residency":"US","country":"United States","inference_location":"US","zdr":{"logging":false,"moderation":false,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"default_on","caching":"unknown","content_storage":"unknown","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"novita/api-us"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp8","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"novita/zai-org/glm-5.3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"zai-org/glm-5.3-flash","capabilities":["text","tools","thinking","vision"],"speed":"fast","cost":0.047,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["low","high","max"],"default":"","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","repetition_penalty","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.15],"output":[0.5],"cached_input":[0.03]},"region":"US","model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"provider":{"slug":"novita","display_name":"Novita","owner_country":"US"},"service_route":{"id":"novita/api-us","adapter":"novita"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"United States","compliance":{"residency":"US","country":"United States","inference_location":"US","zdr":{"logging":false,"moderation":false,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"default_on","caching":"unknown","content_storage":"unknown","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"novita/api-us"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp8","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","file","video"],"output_modalities":["text"]},{"id":"perplexity/glm-5.3","type":"llm","name":"GLM-5.3","model_id":"perplexity/glm-5.3","capabilities":["text","tools","structured_output","thinking","reasoning"],"speed":"medium","cost":0.229,"quality":"great","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1}},"pricing":{"billing_unit":"per_mtok","input":[1.4],"output":[4.4],"cached_input":[0.26]},"region":"GLOBAL","api_type":"responses","model":"glm-5.3","family":"glm-5.3","provider":{"slug":"perplexity","display_name":"Perplexity","owner_country":"US"},"service_route":{"id":"perplexity/agent-global","adapter":"perplexity-agent"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","zdr":{"logging":true,"moderation":null,"caching":true,"training":false,"subprocessors":null},"training":"no","logging":"other","moderation":"unknown","caching":"implicit","content_storage":"retained","dpa_available":null,"third_party_access":null,"transfer_mechanism":"unknown","route_id":"perplexity/agent-global"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"perplexity/glm-5.3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"perplexity/glm-5.3-flash","capabilities":["text","tools","thinking","vision","reasoning","structured_output"],"speed":"fast","cost":0.047,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1}},"pricing":{"billing_unit":"per_mtok","input":[0.15],"output":[0.5],"cached_input":[0.03]},"region":"GLOBAL","api_type":"responses","model":"glm-5.3-flash","family":"glm-5.3-flash","provider":{"slug":"perplexity","display_name":"Perplexity","owner_country":"US"},"service_route":{"id":"perplexity/agent-global","adapter":"perplexity-agent"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","zdr":{"logging":true,"moderation":null,"caching":true,"training":false,"subprocessors":null},"training":"no","logging":"other","moderation":"unknown","caching":"implicit","content_storage":"retained","dpa_available":null,"third_party_access":null,"transfer_mechanism":"unknown","route_id":"perplexity/agent-global"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","file"],"output_modalities":["text"]},{"id":"sference/zai-org/GLM-5.3","type":"llm","name":"GLM-5.3","model_id":"zai-org/GLM-5.3","capabilities":["text","tools","structured_output","thinking","reasoning"],"speed":"medium","cost":0.221,"quality":"great","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["minimal","low","medium","high","xhigh"],"default":"","wire":"reasoning_effort"},"caller_params":["top_k"]},"pricing":{"billing_unit":"per_mtok","input":[1.2],"output":[4.2],"cached_input":[0.26]},"region":"EU","model":"glm-5.3","family":"glm-5.3","pooled":true,"provider":{"slug":"sference","display_name":"Sference","owner_country":"GB"},"service_route":{"id":"sference","adapter":"sference"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"EEA","inference_location":"EEA","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":false},"training":"no","logging":"none","moderation":"none","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":null,"transfer_mechanism":"sccs","route_id":"sference"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"nvfp4","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"sference/zai-org/GLM-5.3-Flash","type":"llm","name":"GLM-5.3-Flash","model_id":"zai-org/GLM-5.3-Flash","capabilities":["text","tools","vision","structured_output","reasoning","thinking"],"speed":"fast","cost":0.057,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["minimal","low","medium","high","xhigh"],"default":"","wire":"reasoning_effort"},"caller_params":["frequency_penalty","min_p","presence_penalty","repetition_penalty","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.2],"output":[0.6],"cached_input":[0.07]},"region":"EU","model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"provider":{"slug":"sference","display_name":"Sference","owner_country":"GB"},"service_route":{"id":"sference","adapter":"sference"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"EEA","inference_location":"EEA","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":false},"training":"no","logging":"none","moderation":"none","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":null,"transfer_mechanism":"sccs","route_id":"sference"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"nvfp4","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image"],"output_modalities":["text"]},{"id":"tencent:sg/glm-5.3","type":"llm","name":"GLM-5.3","model_id":"glm-5.3","capabilities":["text","tools","reasoning","thinking"],"speed":"medium","cost":0.229,"quality":"great","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":1,"default":1},"reasoning":{"supported":["low","high","max"],"default":"","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs"]},"pricing":{"billing_unit":"per_mtok","input":[1.4],"output":[4.4],"cached_input":[0.26]},"region":"GLOBAL","model":"glm-5.3","family":"glm-5.3","pooled":true,"provider":{"slug":"tencent-cloud","display_name":"Tencent Cloud","owner_country":"CN"},"service_route":{"id":"tencent:sg","adapter":"tencent:sg"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"default_on","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"tencent:sg"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"tencent:sg/glm-5.3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"glm-5.3-flash","capabilities":["text","tools","vision","reasoning","thinking"],"speed":"fast","cost":0.047,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":1,"default":1},"reasoning":{"supported":["low","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","seed"]},"pricing":{"billing_unit":"per_mtok","input":[0.15],"output":[0.5],"cached_input":[0.03]},"region":"GLOBAL","model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"provider":{"slug":"tencent-cloud","display_name":"Tencent Cloud","owner_country":"CN"},"service_route":{"id":"tencent:sg","adapter":"tencent:sg"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"default_on","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"tencent:sg"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","file","video"],"output_modalities":["text"]},{"id":"tencent:sg/glm-5.3-flashx","type":"llm","name":"GLM-5.3-FlashX","model_id":"glm-5.3-flashx","capabilities":["text","tools","vision","reasoning","thinking"],"speed":"fast","cost":0.1,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"","params":{"temperature":{"min":0,"max":1,"default":1},"reasoning":{"supported":["low","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs"]},"pricing":{"billing_unit":"per_mtok","input":[0.37],"output":[1.25],"cached_input":[0.075]},"region":"GLOBAL","model":"glm-5.3-flashx","family":"glm-5.3-flashx","pooled":true,"provider":{"slug":"tencent-cloud","display_name":"Tencent Cloud","owner_country":"CN"},"service_route":{"id":"tencent:sg","adapter":"tencent:sg"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"GLOBAL","inference_location":"GLOBAL","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"default_on","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"sccs","route_id":"tencent:sg"},"release":{"date":"2026-09-18","source":"vendor","url":"https://docs.z.ai/guides/vlm/glm-5.3-flash"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"model_card","url":"https://docs.z.ai/guides/vlm/glm-5.3-flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","file","video"],"output_modalities":["text"]},{"id":"tensorx/z-ai/glm-5.3","type":"llm","name":"GLM-5.3","model_id":"z-ai/glm-5.3","capabilities":["text","tools","structured_output","thinking","reasoning"],"speed":"medium","cost":0.239,"quality":"great","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 on TensorX — EU-resident serverless inference (Dublin/Helsinki) with zero data retention","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"max","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","frequency_penalty","presence_penalty","repetition_penalty","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[1.75],"output":[4.5],"cached_input":[0.44]},"region":"EU","model":"glm-5.3","family":"glm-5.3","pooled":true,"provider":{"slug":"tensorx","display_name":"TensorX","owner_country":"IE"},"service_route":{"id":"tensorx","adapter":"tensorx"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"EU","inference_location":"EU","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"none","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":null,"transfer_mechanism":"eu_resident","route_id":"tensorx"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp8","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"tensorx/z-ai/glm-5.3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"z-ai/glm-5.3-flash","capabilities":["text","tools","thinking","vision","reasoning","structured_output"],"speed":"fast","cost":0.05,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","minimal","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","frequency_penalty","presence_penalty","repetition_penalty","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.2],"output":[0.5],"cached_input":[0.05]},"region":"EU","model":"glm-5.3-flash","family":"glm-5.3-flash","provider":{"slug":"tensorx","display_name":"TensorX","owner_country":"IE"},"service_route":{"id":"tensorx","adapter":"tensorx"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"compliance":{"residency":"EU","inference_location":"EU","zdr":{"logging":false,"moderation":false,"caching":true,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"none","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":null,"transfer_mechanism":"eu_resident","route_id":"tensorx"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"quantization":"fp8","references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","video"],"output_modalities":["text"]},{"id":"wafer/GLM-5.3","type":"llm","name":"GLM-5.3","model_id":"GLM-5.3","capabilities":["text","tools","structured_output","thinking","reasoning"],"speed":"medium","cost":0.225,"quality":"great","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 on Wafer - US serverless inference with per-request zero data retention (Wafer-ZDR pinned by the adapter); Z.ai's 391B NVFP4 MoE flagship with the full 1M-token context","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["low","medium","high","xhigh","max"],"default":"low","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","presence_penalty","repetition_penalty","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[1.19],"output":[4.4],"cached_input":[0.26]},"region":"US","model":"glm-5.3","family":"glm-5.3","pooled":true,"provider":{"slug":"wafer","display_name":"Wafer","owner_country":"US"},"service_route":{"id":"wafer","adapter":"wafer"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"United States","compliance":{"residency":"US","country":"United States","inference_location":"US","zdr":{"logging":false,"moderation":false,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"none","caching":"unknown","content_storage":"unknown","dpa_available":true,"third_party_access":null,"transfer_mechanism":"sccs","route_id":"wafer"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"wafer/GLM-5.3-Flash","type":"llm","name":"GLM-5.3-Flash","model_id":"GLM-5.3-Flash","capabilities":["text","tools","thinking","reasoning","structured_output"],"speed":"fast","cost":0.034,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":2,"default":1},"reasoning":{"supported":["none","low","medium","high","xhigh","max"],"default":"","wire":"reasoning_effort"},"caller_params":["chat_template_kwargs","repetition_penalty","top_k"]},"pricing":{"billing_unit":"per_mtok","input":[0.1],"output":[0.35],"cached_input":[0.02]},"region":"US","model":"glm-5.3-flash","family":"glm-5.3-flash","provider":{"slug":"wafer","display_name":"Wafer","owner_country":"US"},"service_route":{"id":"wafer","adapter":"wafer"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"United States","compliance":{"residency":"US","country":"United States","inference_location":"US","zdr":{"logging":false,"moderation":false,"caching":null,"training":false,"subprocessors":null},"training":"no","logging":"none","moderation":"none","caching":"unknown","content_storage":"unknown","dpa_available":true,"third_party_access":null,"transfer_mechanism":"sccs","route_id":"wafer"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image"],"output_modalities":["text"]},{"id":"zai/glm-5.3","type":"llm","name":"GLM-5.3","model_id":"glm-5.3","capabilities":["text","tools","structured_output","reasoning","thinking"],"speed":"medium","cost":0.229,"quality":"great","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3 is Z.ai's flagship coding and agentic model, post-trained on the GLM-5.2 base (753B MoE): 1M-token context, 128K output, text-only, reasoning always on with a low/high/max effort ladder.","params":{"temperature":{"min":0,"max":1,"default":1},"max_tokens":true,"reasoning":{"supported":["low","high","max"],"default":"","wire":"reasoning_effort"},"caller_params":["top_k"]},"pricing":{"billing_unit":"per_mtok","input":[1.4],"output":[4.4],"cached_input":[0.26]},"region":"GLOBAL","model":"glm-5.3","family":"glm-5.3","pooled":true,"access":"open","provider":{"slug":"zai","display_name":"Z.ai","owner_country":"CN"},"service_route":{"id":"zai","adapter":"zai"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"Singapore","compliance":{"residency":"GLOBAL","country":"Singapore","inference_location":"SG","zdr":{"logging":false,"moderation":true,"caching":true,"training":false,"subprocessors":true},"training":"no","logging":"none","moderation":"default_on","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"unknown","route_id":"zai"},"release":{"date":"2026-08-18","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text"],"output_modalities":["text"]},{"id":"zai/glm-5.3-flash","type":"llm","name":"GLM-5.3-Flash","model_id":"glm-5.3-flash","capabilities":["text","tools","vision","reasoning","thinking"],"speed":"fast","cost":0.047,"quality":"good","context_window":1048576,"max_output_tokens":131072,"description":"GLM-5.3-Flash is a natively multimodal model from Z.ai with 320B total / 18B active parameters. Its hybrid sparse-and-linear attention architecture keeps long-context behaviour accurate while cutting compute overhead, and it targets efficient coding and long-horizon agent tasks.","params":{"temperature":{"min":0,"max":1,"default":1},"max_tokens":true,"reasoning":{"supported":["low","high","max"],"default":"","wire":"reasoning_effort"}},"pricing":{"billing_unit":"per_mtok","input":[0.15],"output":[0.5],"cached_input":[0.03]},"region":"GLOBAL","model":"glm-5.3-flash","family":"glm-5.3-flash","pooled":true,"access":"open","provider":{"slug":"zai","display_name":"Z.ai","owner_country":"CN"},"service_route":{"id":"zai","adapter":"zai"},"maker":{"slug":"zai","display_name":"Z.ai","country":"CN"},"country":"Singapore","compliance":{"residency":"GLOBAL","country":"Singapore","inference_location":"SG","zdr":{"logging":false,"moderation":true,"caching":true,"training":false,"subprocessors":true},"training":"no","logging":"none","moderation":"default_on","caching":"implicit","content_storage":"ephemeral","dpa_available":true,"third_party_access":true,"transfer_mechanism":"unknown","route_id":"zai"},"release":{"date":"2026-08-26","source":"vendor","url":"https://docs.z.ai/release-notes/new-released"},"references":[{"kind":"model_card","url":"https://huggingface.co/zai-org/GLM-5.3-Flash","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"},{"kind":"paper","url":"https://arxiv.org/abs/2602.15763","created_at":"2026-10-07T06:54:33.956988Z","updated_at":"2026-10-07T06:54:33.956988Z"}],"input_modalities":["text","image","file"],"output_modalities":["text"]}],"total":47,"offset":0,"limit":50}