Spreadsheet Manipulation and Reasoning on SpreadsheetBench held-out (test)
85ScoreSkillOpt
Evaluation results
| Method | Links | |
|---|---|---|
| SkillOptModel=GPT–5.5, Harness=Codex harness2026.05 | 85 | |
| SkillOptModel=GPT–5.5, Harness=No harness / direct chat2026.05 | 80.7 | |
| SkillOptModel=GPT–5.5, Harness=Claude Code harness2026.05 | 80.4 | |
| EvoSkillModel=GPT–5.5, Harness=Claude Code harness2026.05 | 75 | |
| GEPAModel=GPT–5.5, Harness=No harness / direct chat2026.05 | 73.6 | |
| Human skillModel=GPT–5.5, Harness=No harness / direct chat2026.05 | 72.9 | |
| EvoSkillModel=GPT–5.5, Harness=Codex harness2026.05 | 67.5 | |
| SkillOptModel=GPT–5.4, Harness=No harness / direct chat2026.05 | 62.5 | |
| GEPAModel=GPT–5.4, Harness=No harness / direct chat2026.05 | 61.1 | |
| Human skillModel=GPT–5.4, Harness=No harness / direct chat2026.05 | 59.3 | |
| SkillOptModel=GPT–5.2, Harness=No harness / direct chat2026.05 | 57.1 | |
| GEPAModel=GPT–5.2, Harness=No harness / direct chat2026.05 | 54.3 | |
| Trace2SkillModel=GPT–5.4, Harness=No harness / direct chat2026.05 | 53.2 | |
| Human skillModel=GPT–5.5, Harness=Codex harness2026.05 | 50.7 | |
| Trace2SkillModel=GPT–5.5, Harness=No harness / direct chat2026.05 | 49.6 | |
| Human skillModel=GPT–5.2, Harness=No harness / direct chat2026.05 | 48.2 | |
| SkillOptModel=GPT–5.4-mini, Harness=No harness / direct chat2026.05 | 47.5 | |
| SkillOptModel=Qwen3.6–35B-A3B, Harness=No harness / direct chat2026.05 | 47.5 | |
| LLM skillModel=GPT–5.4, Harness=No harness / direct chat2026.05 | 46.1 | |
| GEPAModel=Qwen3.6–35B-A3B, Harness=No harness / direct chat2026.05 | 45.4 | |
| Human skillModel=Qwen3.6–35B-A3B, Harness=No harness / direct chat2026.05 | 44.3 | |
| Trace2SkillModel=GPT–5.2, Harness=No harness / direct chat2026.05 | 43.9 | |
| LLM skillModel=GPT–5.5, Harness=No harness / direct chat2026.05 | 43.2 | |
| Human skillModel=GPT–5.4-mini, Harness=No harness / direct chat2026.05 | 42.9 | |
| LLM skillModel=Qwen3.6–35B-A3B, Harness=No harness / direct chat2026.05 | 42.9 | |
| GEPAModel=GPT–5.4-mini, Harness=No harness / direct chat2026.05 | 42.5 | |
| SkillOptModel=GPT–5.4-nano, Harness=No harness / direct chat2026.05 | 42.5 | |
| No skillModel=GPT–5.5, Harness=No harness / direct chat2026.05 | 41.8 | |
| Human skillModel=GPT–5.4-nano, Harness=No harness / direct chat2026.05 | 41.8 | |
| No skillModel=GPT–5.4, Harness=No harness / direct chat2026.05 | 41.4 | |
| TextGradModel=GPT–5.5, Harness=No harness / direct chat2026.05 | 41.1 | |
| Trace2SkillModel=GPT–5.4-mini, Harness=No harness / direct chat2026.05 | 40.7 | |
| TextGradModel=GPT–5.4, Harness=No harness / direct chat2026.05 | 38.6 | |
| LLM skillModel=GPT–5.2, Harness=No harness / direct chat2026.05 | 38.6 | |
| TextGradModel=GPT–5.4-mini, Harness=No harness / direct chat2026.05 | 38.2 | |
| No skillModel=GPT–5.2, Harness=No harness / direct chat2026.05 | 38.2 | |
| No skillModel=Qwen3.6–35B-A3B, Harness=No harness / direct chat2026.05 | 38.2 | |
| GEPAModel=GPT–5.4-nano, Harness=No harness / direct chat2026.05 | 37.2 | |
| Human skillModel=GPT–5.5, Harness=Claude Code harness2026.05 | 37.1 | |
| LLM skillModel=GPT–5.5, Harness=Claude Code harness2026.05 | 37.1 | |
| LLM skillModel=GPT–5.4-mini, Harness=No harness / direct chat2026.05 | 36.8 | |
| No skillModel=GPT–5.4-mini, Harness=No harness / direct chat2026.05 | 36.1 | |
| Trace2SkillModel=GPT–5.4-nano, Harness=No harness / direct chat2026.05 | 35.4 | |
| LLM skillModel=GPT–5.4-nano, Harness=No harness / direct chat2026.05 | 33.6 | |
| Trace2SkillModel=Qwen3.6–35B-A3B, Harness=No harness / direct chat2026.05 | 33.2 | |
| TextGradModel=GPT–5.4-nano, Harness=No harness / direct chat2026.05 | 32.5 | |
| TextGradModel=GPT–5.2, Harness=No harness / direct chat2026.05 | 32.1 | |
| No skillModel=GPT–5.5, Harness=Codex harness2026.05 | 27.5 | |
| LLM skillModel=GPT–5.5, Harness=Codex harness2026.05 | 25 | |
| SkillOptModel=Qwen3.5–4B, Harness=No harness / direct chat2026.05 | 23.9 | |
| No skillModel=GPT–5.4-nano, Harness=No harness / direct chat2026.05 | 23.5 | |
| TextGradModel=Qwen3.6–35B-A3B, Harness=No harness / direct chat2026.05 | 22.9 | |
| No skillModel=GPT–5.5, Harness=Claude Code harness2026.05 | 22.1 | |
| Trace2SkillModel=Qwen3.5–4B, Harness=No harness / direct chat2026.05 | 19.3 | |
| GEPAModel=Qwen3.5–4B, Harness=No harness / direct chat2026.05 | 16.9 | |
| Human skillModel=Qwen3.5–4B, Harness=No harness / direct chat2026.05 | 16.4 | |
| TextGradModel=Qwen3.5–4B, Harness=No harness / direct chat2026.05 | 13.9 | |
| LLM skillModel=Qwen3.5–4B, Harness=No harness / direct chat2026.05 | 10.4 | |
| No skillModel=Qwen3.5–4B, Harness=No harness / direct chat2026.05 | 9.3 |