Doc2Harness - Direct drive: leaderboard
Table 2; three independent builds per model in the same fixed builder harness; development document plus interface skeleton; 328 end-to-end test cases; frozen artifact tested by scripted mock model; category means and overall weighted by test count. Metric: Pass rate (%). Higher is better. Source: https://arxiv.org/html/2610.20344v1
Metric: Pass rate (%). Source: arxiv.org. Saturation forecast: Around 2029. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra | 39.4 | #5 |
| 2 | GPT-5.6 Sol | 32.3 | #18 |
| 3 | Claude Opus 4.6 | 25.3 | #60 |
| 4 | GPT-5.5 | 24.2 | #32 |
| 5 | DeepSeek V4 Pro | 23.2 | #99 |
| 6 | Grok 4.7 | 22.2 | #27 |
| 7 | Kimi K3 | 18.2 | #30 |
| 8 | GLM-5.3 | 18.2 | #40 |
| 9 | Claude Opus 5 | 17.2 | #12 |
| 10 | Qwen 3.8 Max (0902) | 15.2 | #28 |
| 11 | DeepSeek V4.1 Flash | 14.1 | #36 |
| 12 | MiMo-V2.6-Flash | 14.1 | #62 |
| 13 | Step 5 Preview | 14.1 | #41 |
| 14 | Hy3 | 13.1 | #90 |
| 15 | Claude Opus 4.7 | 12.1 | #49 |
No result here: #3 Claude Opus 5.5, #7 Gemini 4 Argon, #8 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=doc2harness-direct-drive · How It Works · Data refreshed daily, snapshot 2026-10-16.