FORTE - Pass@3: leaderboard
FORTE office-productivity agent benchmark (AGI-Eval): Pass@3 is the share of tasks solved in at least one of three independent OpenClaw runs, each graded all-or-nothing against expert rubrics by an LLM judge, with thinking enabled at the highest available effort. %, higher is better.
Metric: Pass@3 (%, task solved in at least one of three runs). Source: agi-eval-official.github.io. Saturation forecast: Estimated already saturated. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 85.25 | #33 |
| 2 | GLM-5.1 | 85.25 | #109 |
| 3 | GLM-5.2 | 85.25 | #80 |
| 4 | Claude Opus 4.8 | 84.7 | #39 |
| 5 | LongCat 2.0 | 84.7 | #220 |
| 6 | Claude Opus 4.7 | 83.61 | #51 |
| 7 | DeepSeek V4 Pro | 83.06 | #112 |
| 8 | MiniMax-M3 | 82.51 | #151 |
| 9 | Gemini 3.1 Pro (Preview) | 81.97 | #64 |
| 10 | DeepSeek V4 Flash | 81.96 | #155 |
| 11 | Claude Opus 4.6 | 80.87 | #67 |
| 12 | Kimi K2.6 | 80.33 | #113 |
| 13 | GPT-5.4 | 79.23 | #83 |
| 14 | Claude Sonnet 4.6 | 78.14 | #94 |
| 15 | Qwen 3.6 Plus | 75.96 | #124 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=forte-pass-3 · How It Works · Data refreshed daily, snapshot 2026-11-05.