Benchmark
PostTrainBench
Measures a model's ability to run post-training itself. Hardware differs between reported runs (H100 vs H20), which moves the score.
- Released
- 2026-01-20
- Categories
- ai_research
- Openness
- unknown
- Source
- model_reports
- Reported scores
- 3
Reported scores
model_reports
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| GLM-5.2 | Z.ai | 34.3 | 2026-06-16 | evidence |
| Hy4 preview | Tencent | 35.6 | 2026-08-28 | evidence |
| Kimi K3 | Moonshot AI | 36.6 | 2026-06-13 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.