§ Ranked #11 by discriminative power
SWE-Bench-Pro.
An environment for software engineering (audited). Across 4 models with public results it spreads the best and worst 11% — it still sorts frontier models, so training on it can still move yours.
§ Public model scores
Who wins SWE-Bench-Pro.
Best public result per model entry, normalized 0..1. The spread between the top and bottom rows is what makes this environment worth — or not worth — a training run.
| # | Model | resolve-rate |
|---|---|---|
| 01 | Blitzy | 67% |
| 02 | WarpGrep-v2 | 59% |
| 03 | GPT-5.4 | 58% |
| 04 | ClaudeCode-Opus-4.5 | 55% |
§ Nearby in the ranking
| # | Environment | Spread | Discriminative |
|---|---|---|---|
| 09 | Cua-BenchGUI computer-use (desktop + mobile) | 14% | 0.14 |
| 10 | CompileBenchcompile/cross-compile real OSS | 27% | 0.14 |
| 11 | SWE-Bench-Prosoftware engineering (audited) | 11% | 0.11 |
| 12 | COBOLBenchlegacy enterprise (COBOL) — floor effect | 2% | 0.02 |
§ Work with us
Need one that still separates models?
When the public environment for your capability saturates, you can’t tell your models apart and you can’t train past it. We build private, contamination-resistant, verifiable-reward environments and evals on a hold-out set — designed to discriminate where the public ones no longer do.