GLM-4.7 is a new 358-billion-parameter AI model with open weights released by the Beijing-based company Z.ai.

How does it perform on coding tasks?

It tops open-source benchmarks for coding, achieving a score of 73.8% on SWE-bench Verified and ranking first on LMArena's WebDev leaderboard.

What are its reasoning capabilities?

The model demonstrated strong reasoning skills, scoring 85.7% on the GPQA-Diamond benchmark.

How can I use GLM-4.7?

You can access the model through Z.ai's API, download the weights from Hugging Face, or sign up for a new coding plan for $3 per month.

What is the context window size for this model?

GLM-4.7 supports a large context window of 128K tokens.

GLM-4.7

Top open-source AI model for coding and reasoning.

From Zhipu

Text Generator

22 Dec 2025

10K+

Freemium

Pricing Model

Freemium

Starting Price

$3/Month

About This AI

GLM-4.7 is a 358-billion-parameter open-weights AI model from Beijing-based Z.ai. It currently leads open-source benchmarks in coding, reasoning, and agentic tasks.

The model supports a massive 128K context window and is noted for its speed on powerful consumer hardware like the M3 Ultra Mac. It highlights China’s growing strength in frontier AI.

Chatbot Coding Conversation LLM Text To Image

Pricing

Pricing Model

Freemium

Starting Price

$3/Month

Key Features

358 Billion Parameters: A massive scale model with open weights.
Leads in SWE-bench Verified (coding), GPQA-Diamond (reasoning), and HLE with tools (agentic).
Capable of processing extensive inputs.
Demonstrated generation speeds of 16 tokens/second on M3 Ultra for complex tasks like game creation.
Ranks first among open models on LMArena's WebDev leaderboard.

Pros

Best-in-class open-source performance for coding and reasoning.
Large context window for handling complex projects.
Multiple affordable access points, including a $3/month plan and open weights.
Runs efficiently on high-end local hardware.

Cons

No specific cons were mentioned in the source text.

Best For Software development, complex reasoning tasks, building AI agents, and web development projects.

FAQs

What is GLM-4.7?
GLM-4.7 is a new 358-billion-parameter AI model with open weights released by the Beijing-based company Z.ai.
How does it perform on coding tasks?
It tops open-source benchmarks for coding, achieving a score of 73.8% on SWE-bench Verified and ranking first on LMArena’s WebDev leaderboard.
What are its reasoning capabilities?
The model demonstrated strong reasoning skills, scoring 85.7% on the GPQA-Diamond benchmark.
How can I use GLM-4.7?
You can access the model through Z.ai’s API, download the weights from Hugging Face, or sign up for a new coding plan for $3 per month.
What is the context window size for this model?
GLM-4.7 supports a large context window of 128K tokens.