LMGame-Bench Ace Attorney: leaderboard
GamingAgent chooses dialogue/evidence actions in legal adventure cases; native leaderboard correct-action progression score. Primary GamingAgent harnessed cohort, distinct from model-only tab; full published score precision and source run annotations retained.
Metric: Score. Source: huggingface.co. Saturation forecast: Around November 2026. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 (2025-04-16) | 16 | #121 |
| 2 | O1 (2024-12-17) | 16 | #149 |
| 3 | GPT-5 (High) | 9 | #98 (GPT-5) |
| 4 | Claude 3.7 Sonnet (20250219) | 7 | #201 |
| 5 | Gemini 2.5 Pro (Preview 05-06) | 7 | #147 |
| 6 | Claude Opus 4 (20250514) | 6 | #154 |
| 7 | O4 Mini (2025-04-16) | 4 | #179 |
| 8 | Gemini 2.5 Flash (Preview 04-17) | 4 | #237 |
| 9 | Claude Sonnet 4 (20250514) | 3.67 | #213 |
| 10 | GPT-4.1 | 2 | #248 |
| 11 | Claude 3.5 Sonnet (20241022) | 2 | #291 |
| 12 | DeepSeek R1 | 0 | #251 |
| 13 | Llama 4 Maverick Instruct FP8 | 0 | #414 |
| 14 | GPT-4o (2024-11-20) | 0 | #381 |
| 15 | O1 Mini (2024-09-12) | 0 | #403 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=lmgame-bench-ace-attorney · How It Works · Data refreshed daily, snapshot 2026-10-20.