---
id: "z-ai/glm-5.3"
also: ["z-ai/glm-5.3-20260816", "zai-org/GLM-5.3", "hf/zai-org/GLM-5.3", "requesty/zai/glm-5.3", "vercel/zai/glm-5.3"]
card_usd_per_m: "1.4 in, 4.4 out, at z-ai/fp8"
cheapest_list_usd_per_m: "0.0699 in, 1.4 out, at wafer/us on openrouter"
cheapest_effective_usd_per_m: "0.069 in, 1.3996 out, at wafer/us on openrouter"
context: 1048576
max_output: 943718
input: ["text"]
output: ["text"]
released: "2026-08-18"
reasoning: ["max", "high", "low"]
api: ["/api/v1/chat/completions", "/api/v1/completions", "/api/v1/messages", "/api/v1/responses"]
params: ["frequency_penalty", "include_reasoning", "logit_bias", "logprobs", "max_tokens", "min_p", "parallel_tool_calls", "presence_penalty", "reasoning", "reasoning_effort", "repetition_penalty", "response_format", "seed", "stop", "structured_outputs", "temperature", "tool_choice", "tools", "top_k", "top_logprobs", "top_p"]
endpoints: 110
gateways: "openrouter (57), vercel (34), requesty (11), hf (8)"
no_training: 80
trains: 14
retains_prompts: 26
intelligence: 31.7000
coding: 74.8000
agentic: 53.1000
---
# Z.ai: GLM 5.3

GLM 5.3 by Z.ai is an open weights model served by Mistral on EU infrastructure with no Mistral modifications. 1M token context window, 128K max output. Text only. Supports reasoning, function calling, structured outputs, predicted outputs and prefix. Built for long context coding and agentic workflows.