GLM-4.7 (Thinking): benchmark results
Provider: Zhipu. Released 2025-12-22. Access: Open.
Aggregate Rating 1661 ± 13, rank #488 of 2168 rated models, from 68 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| General365 - Recursive & Backtracking | 68.8 | Accuracy (%) on the problems labelled Recursive & Backtracki | 96 |
| HeaRTS - Future Forecasting | 68 | Mean task score (x100) over HeaRTS's future forecasting task | 93.3 |
| HeaRTS - Physiological Classification | 45 | Mean task score (x100) over HeaRTS's physiological classific | 93.3 |
| HPR-Bench - Tag Inference - Lifestyle Indicators | 70.16 | Avg@10 Tag Accuracy (%; mean of 10 samples) | 92.9 |
| General365 - Optimal Strategy | 49.2 | Accuracy (%) on the problems labelled Optimal Strategy (a pr | 88 |
| General365 - Probability & Uncertainty | 58 | Accuracy (%) on the problems labelled Probability & Uncertai | 88 |
| BenchTable - STEM | 77.2 | Weighted Score (%) | 87.9 |
| HeaRTS | 66 | Overall score (x100): macro-average of the Perception, Infer | 86.7 |
| HeaRTS - Event Localization | 38 | Mean task score (x100) over HeaRTS's event localization task | 86.7 |
| BenchTable - Reasoning | 71.6 | Weighted Score (%) | 86.3 |
| LLM2014 Logic 2025-12 | 65.49 | Median Score | 86 |
| BenchTable | 69.6 | Total Score (%) | 85.5 |
Interactive version: theaggregate.ai/model?slug=glm-4-7-thinking · How It Works · Data refreshed daily, snapshot 2026-10-20.