Orak Battle Arena - Street Fighter III: leaderboard
Orak Battle Arena for Street Fighter III: zero-shot LLM agents fight each other in pairwise matches, both playing Ken, and the results are summarized as Elo ratings (higher is better). Adversarial play reorders models relative to the single-player scores; seven agents, so ratings rest on few matches. From KRAFTON AI's Orak benchmark.
Metric: Elo. Source: krafton-ai.github.io. Saturation forecast: Estimated already saturated. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o (2024-08-06) | 1602.9 | #343 |
| 2 | O3 Mini (2025-01-31) | 1602.9 | #249 |
| 3 | GPT-4o Mini | 1453 | #598 |
| 4 | Claude 3.7 Sonnet | 1434.3 | #259 |
| 5 | Qwen 2.5 7B Instruct | 1415.4 | #863 |
| 6 | Llama 3.2 3B Instruct | 1358.4 | #1327 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=orak-battle-arena-street-fighter-iii · How It Works · Data refreshed daily, snapshot 2026-11-05.