GLM-4.7 represents Zhipu AI's latest advancement in large language models, prioritizing agentic coding, stable multi-step reasoning, and complex workflows. Released in December 2025, it supports a 200K token context window and up to 128K output tokens, enabling robust handling of extended tasks.
GLM-4.7 leads open-source models in coding benchmarks at 128K context, achieving 73.8% on SWE-bench Verified (up 5.8% from GLM-4.6) and 84.9% on LiveCodeBench v6, surpassing Claude Sonnet 4.5. Reasoning scores include 42.8% on HLE with tools (41% gain over predecessor) and 95.7% on AIME 2025.
GLM-4.7 excels in programming with a "think before acting" mechanism, improving multi-language code generation and terminal agent performance. It integrates function calling, structured JSON outputs, context caching, and real-time streaming for seamless developer workflows.
Seamlessly writes, debugs, and deploys code across languages—ideal for terminal-based and IDE-integrated workflows.
Generates modern, aesthetically refined HTML/CSS/JS and presentation slides with accurate layout, spacing, and design sensibility.
Creates high-fidelity PPTs, posters, technical documentation, and narrative content with consistent tone and structure.
Excels at synthesizing information from long contexts, extracting insights, and generating evidence-backed summaries.
GLM-4.7 powers agentic coding by autonomously handling requirement decomposition, multi-stack integration, and executable frameworks for prototypes. Developers use it for multimodal apps with real-time controls and gesture recognition, accelerating from concept to deployment.
vs. GLM-4.6: GLM-4.7 boosts coding stability with 73.8% on SWE-bench Verified (up 5.8%) and 41% on Terminal Bench 2.0 (up 16.5%), plus 41% HLE gain via preserved thinking.
vs. Claude Sonnet 4.5: Matches coding prowess at 84.9% LiveCodeBench v6 (outpacing it) and leads open-source on SWE-bench, but trails slightly in Terminal Bench; excels in tool use at 87.4% τ²-Bench.
vs. GPT-5.1: Competitive at 84.9% LiveCodeBench v6 and 73.8% SWE-bench Verified, with superior open-source accessibility and 3x speed in real tests
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "zhipu/glm-4.7", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "zhipu/glm-4.7", "messages": [ { "role": "user", "content": "Hello!" } ] }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"zhipu/glm-4.7","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Cached input |
| Benchmark | Score | What it measures | Source | Retrieved |
|---|---|---|---|---|
| LiveCodeBench | 84.9% | Contamination-free competitive programming problems | Source | July 12, 2026 |
| SWE-bench Verified | 73.8% | Resolving verified real GitHub issues | Source | July 12, 2026 |
| GPQA Diamond | 85.7% | Google-proof graduate science questions (hardest subset) | Source | July 12, 2026 |
| AIME 2025 | 95.7% | Competition mathematics (AIME), 2025 | Source | July 12, 2026 |
| Intelligence | 22.2 | Composite score across standardised reasoning, knowledge and problem-solving evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Coding | 45.3 | Composite score across standardised coding evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Math | 95 | Composite score across standardised mathematics evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
GLM-4.7 This page | Coding + agents | |||
| Reasoning + agents | ||||
| Balanced coding + agents | ||||
| Long-context, multimodal & agentic workflows | ||||
| Reasoning + agents |
GLM-4.7 has a 200,000 tokens context window and can return up to 128,000 tokens.
GLM-4.7 takes text as input and returns text.
Use zhipu/glm-4.7 as the model id. Requests go to https://api.aimlapi.com/v1/chat/completions.
GLM-4.7 became available on December 23, 2025.
GLM-4.7 is priced at input $0.78 / 1M tokens, output $2.86 / 1M tokens, cached input $0.143 / 1M tokens.
Yes, GLM-4.7 can stream responses as they are generated.
GLM-4.7 was built by Zhipu AI.
