SOTA REASONING MODEL

GLM 4.7

The next evolution in agentic intelligence. Featuring interleaved thinking and preserved reasoning across complex workflows.

95.7% AIME 25 Math
200K Context Window
87.4% τ²-Bench Agentic

Performance Analysis

Benchmark GLM 4.7 Claude Sonnet 4.5 Gemini 3.0
Math (AIME 25) 95.7% 32.0% 95.0%
Coding (LCB) 84.9% 64.0% 78.2%
Agentic (τ²-Bench) 87.4% 87.2% 82.1%
Logic (HLE) 42.8% 32.0% 38.5%
// System Initialization...
> Analyzing multi-step request at turn 4...
> Interleaving tool usage with logical verification...
> Found optimal path: 98.2% confidence interval.
[Reasoning Preserved] GLM 4.7 reuses this context for subsequent agentic steps.
NEW YEAR 2025 SPECIAL

Start Building Today

Get 50% OFF your first purchase + an extra 10% OFF with invite code. Automatic discount applied via link below.

Claim New Year Discount celebration View Raw Data

Official TRAE IDE Support

GLM 4.7 is now officially supported on TRAE IDE through the Z.AI Plan. Get access to powerful custom agents developed for the TRAE community.

View TRAE Agents extension