## Stet benchmark evidence

Inspection only. 15 setups on 24 tasks from Stet-AI/stet.

Report SHA-256: 0bbc1f6210cf4ed08f9a425024a6e7e45e25a003fc5fe3585536c3c694d96f46
Accepted head: sha256:d34123b5df1396ebc24f274af135704f492fee23dd2f05ba6770e7a6d0970661
Accepted at: 2026-10-21T18:23:38.660Z
Report generated at: 2026-10-20T21:49:46Z
Decision: inconclusive; recommendation: inspect; confidence: low.

Measure on your own repository before adopting a setup.

- SWE-2 Max: Stet score 66.7; cost per task not recorded; agent seconds per task 4095.4.
- GPT 6.1 Sol medium: Stet score 65.8; cost per task $1.10; agent seconds per task 706.6.
- Opus 5: Stet score 65.7; cost per task $6.78; agent seconds per task 1392.2.
- Grok 4.7 High: Stet score 64.0; cost per task $5.76; agent seconds per task 1135.3.
- Pareto 26.9: Stet score 60.0; cost per task $12.47; agent seconds per task 2351.1.
- Haiku 5.5 max: Stet score 58.9; cost per task $3.85; agent seconds per task 3504.1.
- GPT 6 Sol medium: Stet score 58.5; cost per task $3.01; agent seconds per task 925.1.
- GPT 6 Luna max: Stet score 50.8; cost per task $0.37; agent seconds per task 3583.7.
- DeepSeek V4.1 Flash: Stet score 48.1; cost per task $0.49; agent seconds per task 2956.4.
- Haiku 5.5 low+guidance: Stet score 42.4; cost per task $0.29; agent seconds per task 743.3.
- Haiku 5.5 low: Stet score 41.9; cost per task $0.25; agent seconds per task 592.7.
- Opus 4.8: Stet score 40.8; cost per task $7.05; agent seconds per task 1398.5.
- Muse control: Stet score 34.1; cost per task $0.10; agent seconds per task 793.6.
- GLM 5.3 Flash: Stet score 31.1; cost per task $0.41; agent seconds per task 1604.9.
- GPT 6 Luna low: Stet score 21.2; cost per task $0.05; agent seconds per task 413.5.

Public evidence: /leaderboard/benchmark.json?report=0bbc1f6210cf4ed08f9a425024a6e7e45e25a003fc5fe3585536c3c694d96f46