Model Details
GLM-4.7
Quick answer: GLM-4.7 is Zhipu AI's June 2025 model scoring 52.0% BrowseComp, 42.8% HLE, 85.7% GPQA Diamond, and 23.5% CyberGym. Proprietary, preceding the GLM-5.2 generation.
At a Glance
Where GLM-4.7 leads
- 52.0% BrowseComp — competitive web research
- 42.8% HLE — moderate frontier reasoning
- 85.7% GPQA Diamond — strong graduate-level science
- Zhipu AI ecosystem integration
Where it lags
- Proprietary: no open weights
- Superseded by GLM-5.2 (54.7% HLE, 91.2% GPQA Diamond, 77.0% ARC-AGI)
- 23.5% CyberGym — limited cybersecurity task performance
Best for: Zhipu AI (BigModel) platform users; Chinese-language applications before GLM-5.2 availability; moderate BrowseComp pipelines.
What GLM-4.7 Is
GLM-4.7 is Zhipu AI's June 2025 model in the GLM-4 generation — a stepping stone before the GLM-5 series. The 52.0% BrowseComp and 42.8% HLE represent competitive but not frontier-leading scores at the June 2025 timeframe.
GLM-4.7 is now superseded by GLM-5.2 (July 2026), which substantially improves on all metrics. For new deployments, GLM-5.2 is the recommended GLM model.
Specifications
| Field | Value |
|---|---|
| Organization | Zhipu AI |
| License | Proprietary |
| Release date | June 2025 |
| Modality | Text only |
Pricing
Available via Zhipu AI open platform (bigmodel.cn). Refer to Zhipu pricing.
Public Benchmark Scores
| Benchmark | Score | Source | Date |
|---|---|---|---|
| BrowseComp | 52.0% | Benchgen evaluation | 2025-06 |
| Humanity's Last Exam | 42.8% | Benchgen evaluation | 2025-06 |
| GPQA Diamond | 85.7% | Benchgen evaluation | 2025-06 |
| CyberGym | 23.5% | Benchgen evaluation | 2025-06 |
GLM-4.7 vs Alternatives
| Model | HLE | BrowseComp | GPQA Diamond | License |
|---|---|---|---|---|
| GLM-4.7 | 42.8% | 52.0% | 85.7% | Proprietary |
| GLM-5.2 | 54.7% | — | 91.2% | Proprietary |
| Kimi K2.6 | 36.4% | 83.2% | 90.5% | Apache 2.0 |
GLM-4.7 → GLM-5.2: all metrics improve. Use GLM-5.2 for new deployments unless specific API/pricing constraints favour GLM-4.7.
Related Models & Benchmarks
- Compare: GLM-5.2 · Kimi K2.6
- Benchmarks: BrowseComp · Humanity's Last Exam · GPQA Diamond
Frequently Asked Questions
What is GLM-4.7?
Zhipu AI's June 2025 model scoring 52.0% BrowseComp, 42.8% HLE, 85.7% GPQA Diamond. Proprietary — superseded by GLM-5.2.Specs from Zhipu AI's GLM-4.7 release (June 2025) and Benchgen evaluations. Last updated 2026-07-24.