Orak - Ace Attorney: leaderboard
Orak Ace Attorney: an LLM agent plays courtroom trials, pressing witnesses and presenting evidence; score reflects response correctness and steps taken, normalized to 0–100 and averaged over trials with the game's default agentic strategy. Higher is better. From KRAFTON AI's Orak benchmark (NeurIPS 2025).
Metric: Game Score (0-100). Source: krafton-ai.github.io. Saturation forecast: Estimated already saturated. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 Mini (2025-01-31) | 91.7 | #249 |
| 2 | GPT-4o (2024-08-06) | 85.3 | #343 |
| 3 | DeepSeek R1 | 83.3 | #262 |
| 4 | Claude 3.7 Sonnet | 81.9 | #259 |
| 5 | Gemini 2.5 Pro (Preview 03-25) | 55.7 | #126 |
| 6 | GPT-4o Mini | 28.4 | #598 |
| 7 | Qwen 2.5 3B Instruct | 20 | #1159 |
| 8 | Qwen 2.5 7B Instruct | 9.3 | #863 |
| 9 | Llama 3.2 3B Instruct | 4.6 | #1327 |
| 10 | Llama 3.2 1B Instruct | 1.3 | #1518 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=orak-ace-attorney · How It Works · Data refreshed daily, snapshot 2026-11-05.