GLM-5 FP8: benchmark results
GLM-5 evaluated in FP8 quantization. Provider: Zhipu. Released 2026-02-12. Access: Open.
Aggregate Rating 1732 ± 15, rank #143 of 1656 rated models, from 56 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| One Word, Different Action - Atomic Accuracy | 99.4 | Executable-decision accuracy (%; exact match of the executab | 100 |
| One Word, Different Action - Compositional Accuracy | 94.9 | Executable-decision accuracy (%; exact match of the executab | 100 |
| One Word, Different Action - Decision Invariance | 96.1 | Decision invariance (%; exact match of the executable decisi | 100 |
| One Word, Different Action - Decision Sensitivity | 84.3 | Decision sensitivity (%; exact match of the executable decis | 100 |
| TuRTLe Code Completion (Icarus Verilog) | 83.98 | Aggregated Score (imported) | 100 |
| TuRTLe Code Completion (Verilator) | 82.23 | Aggregated Score (imported) | 100 |
| TurkBench - Math | 81 | Accuracy (%) | 100 |
| TurkBench - STS | 93.55 | Pearson correlation x100 | 100 |
| TurkBench - Turkish General Knowledge | 85.94 | Accuracy (%) | 100 |
| TuRTLe - Icarus Area | 76.97 | Average Score (%) | 97.7 |
| TuRTLe - Icarus Functionality | 76.47 | Average Score (%) | 97.7 |
| TuRTLe - Icarus Power | 75.93 | Average Score (%) | 97.7 |
Interactive version: theaggregate.ai/model?slug=glm-5-fp8 · How It Works · Data refreshed daily, snapshot 2026-10-20.