Review-model evals
How code-review models do on a fixed suite of review cases, so you can pick models for your own review phase. Every comparison is taken over the cases all the chosen models have reviews for.
8 shared cases, 160 reviews (per model: claude-opus-5-5 8, commandcode/ornith/ornith-1.5-122b:free 8, opencode-go/beam-free 8, opencode-go/deepseek-v4.1-flash 8, opencode-go/glm-5.3-flash 24, opencode-go/gpt-6-luna 24, opencode-go/gpt-6-luna (core+lens:concurrency) 8, opencode-go/gpt-6-luna (core+lens:correctness) 8, opencode-go/gpt-6-luna (core+lens:evidence) 8, opencode-go/gpt-6-luna (core+lens:scale) 8, opencode-go/gpt-6-luna (core+lens:seams) 8, opencode-go/gpt-6-luna (core+lens:security) 8, opencode-go/longcat-2.5-preview-free 8, opencode-go/mimo-v2.6-flash 8, opencode-go/minimax-m3.1-flash-free 8, opencode-go/qwen3.8-flash 8) useful
| Model | Weighted recall | MUST recall | False MUSTs (negatives) | Precision | Failed reviews | Cost / review | Wall / review |
|---|---|---|---|---|---|---|---|
| claude-opus-5-5 | 65.4% | 85.7% | 0.00 | 84.4% | 0 | $0.865 | 140s |
| commandcode/ornith/ornith-1.5-122b:free | 31.4% | 14.3% | 0.00 | 83.3% | 0 | $0.000 | 976s |
| opencode-go/beam-free | 40.4% | 28.6% | 0.00 | 100.0% | 0 | $0.001 | 509s |
| opencode-go/deepseek-v4.1-flash | 45.7% | 57.1% | 0.00 | 77.4% | 0 | $0.068 | 495s |
| opencode-go/glm-5.3-flash | 50.8% | 54.8% | 0.00 | 67.7% | 0 | $0.112 | 380s |
| opencode-go/gpt-6-luna | 46.3% | 69.0% | 0.00 | 90.3% | 0 | $0.021 | 87s |
| opencode-go/gpt-6-luna (core+lens:concurrency) | 7.1% | 4.8% | 0.00 | 100.0% | 0 | $0.017 | 65s |
| opencode-go/gpt-6-luna (core+lens:correctness) | 40.1% | 59.5% | 0.00 | 64.3% | 0 | $0.019 | 77s |
| opencode-go/gpt-6-luna (core+lens:evidence) | 36.0% | 52.4% | 0.00 | 71.4% | 0 | $0.017 | 73s |
| opencode-go/gpt-6-luna (core+lens:scale) | 3.2% | 0.0% | 0.00 | 25.0% | 0 | $0.013 | 51s |
| opencode-go/gpt-6-luna (core+lens:seams) | 36.5% | 38.1% | 0.00 | 81.3% | 0 | $0.023 | 102s |
| opencode-go/gpt-6-luna (core+lens:security) | 0.0% | 0.0% | 0.00 | 0.0% | 0 | $0.019 | 74s |
| opencode-go/longcat-2.5-preview-free | 40.5% | 59.5% | 0.00 | 83.3% | 1 | $0.001 | 1085s |
| opencode-go/mimo-v2.6-flash | 35.1% | 42.9% | 0.00 | 100.0% | 4 | $0.049 | 1500s |
| opencode-go/minimax-m3.1-flash-free | 60.2% | 66.7% | 0.00 | 76.2% | 0 | $0.001 | 887s |
| opencode-go/qwen3.8-flash | 57.9% | 57.1% | 0.00 | 76.4% | 1 | $0.102 | 987s |