GLM-4.7 Flash: benchmark results
Zhipu AI's GLM-4.7 Flash, a compact 30B MoE (3B active) speed-tier model (January 2026). Provider: Zhipu. Released 2026-01-19. Access: Open.
Aggregate Rating 1580 ± 6, rank #509 of 1656 rated models, from 784 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SEQUOR | -4.5 | Single constraint: last-minus-first-turn accuracy change (pe | 100 |
| SEQUOR Tuples | -16.5 | Initial constraint tuple: last-minus-first-turn accuracy cha | 100 |
| Tau2-Bench Telecom | 98.83 | Success Rate (%) | 99.8 |
| AA TAU-2 Bench | 98.83 | Accuracy (%) | 99.5 |
| EuroEval Catalan NLU - Guia CAT | 76.04 | Matthews correlation coefficient (%) | 98.8 |
| EuroEval Italian NLU | 66.72 | NLU Average Score (%) | 96 |
| EuroEval Polish NLU - Polemo2 | 95.48 | Matthews correlation coefficient (%) | 95.7 |
| EuroEval Italian NLU - ScaLA IT | 55.95 | Matthews correlation coefficient (%) | 94.7 |
| EuroEval Portuguese NLU - SST-2 PT | 85.45 | Matthews correlation coefficient (%) | 94.3 |
| EuroEval English NLU | 73.87 | NLU Average Score (%) | 94.2 |
| EuroEval Danish NLU - Angry Tweets | 58.48 | Matthews correlation coefficient (%) | 94 |
| EuroEval German NLU - GermEval | 71.44 | Micro F1 excluding MISC (%) | 93.8 |
Interactive version: theaggregate.ai/model?slug=glm-4-7-flash · How It Works · Data refreshed daily, snapshot 2026-10-20.