GLM-5.3 (glm-5.3) Model Details

Overview

1M ContextCode generationAgentic tasksReasoning

GLM-5.3 is the strongest open-source model for coding released by Zhipu. It shares the base model and parameter scale of GLM-5.2 (743B total parameters), with every capability gain coming from extreme post-training scaling. It breaks new ground on complex software engineering, terminal operations and long-horizon agent tasks, with coding capability 50% higher than GLM-5.2; it also fills in cybersecurity, performing strongly on defensive security tasks such as white-box code review and vulnerability discovery. Combined with a million-token context, tiered thinking effort, function calling and MCP tools, it can drive project-scale software development, security audits and complex automation pipelines.

Million-token context

A 1M-token context window loads project-scale codebases and very long documents in a single pass.

Coding leap

Coding capability is 50% higher than GLM-5.2, ranking first among open-source models on complex software engineering and terminal operations.

Cybersecurity

White-box code review and vulnerability discovery, with strong results on defensive security tasks.

Post-training scaling

The IndexShare / SAO / Slime frameworks keep raising the intelligence ceiling on the same base model.

Features

Reasoning✓

Streaming✓

Tool calling✓

Context caching✓

Structured output✓

Web search✓

Batch✕

Prefix completion✕

Fine-tuning✕

Pricing

Input$1.4/M tokens

Input (cache hit)$0.26/M tokens

Output$4.4/M tokens

Rate Limits & Context

Context Window1M

Max Output128K

RPM (requests/min)200

TPM (tokens/min)3M

Built-in Tools

`code_interpreter`

Responses API

`web_extractor`

Responses API

`web_search`

Responses API

API Reference

POSThttps://api.tokenfab.com/v1/chat/completions

Request Example

curl https://api.tokenfab.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $TOKENFAB_API_KEY" \
  -d '{
    "model": "glm-5.3",
    "thinking": { "type": "enabled" },
    "reasoning_effort": "max",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Review the following Python code for security vulnerabilities and suggest fixes."}
    ]
  }'

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.getenv("TOKENFAB_API_KEY"),
    base_url="https://api.tokenfab.com/v1"
)

response = client.chat.completions.create(
    model="glm-5.3",
    thinking={"type": "enabled"},
    reasoning_effort="max",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Review the following Python code for security vulnerabilities and suggest fixes."}
    ]
)

print(response.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.TOKENFAB_API_KEY,
  baseURL: "https://api.tokenfab.com/v1"
});

const response = await client.chat.completions.create({
  model: "glm-5.3",
  thinking: { type: "enabled" },
  reasoning_effort: "max",
  messages: [
    { role: "system", content: "You are a helpful assistant." },
    { role: "user", content: "Review the following Python code for security vulnerabilities and suggest fixes." }
  ]
});

console.log(response.choices[0].message.content);

Response Example

{
  "id": "chatcmpl-glm53-9d2k7f1",
  "object": "chat.completion",
  "created": 1784950000,
  "model": "glm-5.3",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "The review found two risks: 1) user input is not parameterized, which allows SQL injection; 2) sensitive data is written to logs in plain text. Use prepared statements and mask sensitive data before logging."
    },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 42,
    "completion_tokens": 156,
    "total_tokens": 198
  }
}

Request Parameters

Parameter	Type	Required	Description
`model`	string	✓	Model ID, set to glm-5.3
`messages`	array	✓	List of chat messages, each with role and content
`thinking`	object	—	Thinking configuration, {"type":"enabled"} (enabling only; disabling is not supported)
`reasoning_effort`	enum	—	Thinking effort: low / high / max, default max
`temperature`	float	—	Sampling temperature, 0–2, default 0.7
`top_p`	float	—	Nucleus sampling, 0–1, default 0.9
`max_tokens`	int	—	Maximum output tokens
`stream`	bool	—	Whether to stream the response, default false
`tools`	array	—	Tool definitions for function calling
`tool_choice`	string	—	Tool choice policy: auto / none / a named function