Benchmark
ProgramBench
Cleanroom program-rebuild tasks; most models score low, so small differences sit within run-to-run noise.
- Released
- 2026-05-05
- Categories
- coding
- Openness
- unknown
- Source
- model_reports
- Reported scores
- 1
Reported scores
model_reports
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| Hy4 preview | Tencent | 17.5 | 2026-08-28 | 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.