STT-Arena - Easy: leaderboard
Metric: Pass@1 (%) on the 71 easy tasks (a single isolated conflict needing one corrective action); a task passes when every rule-based check function holds in the final state; the model acts in an executable stateful environment whose state changes under scheduled spatio-temporal triggers, with a passive Qwen3.5-397B user simulator and at most 50 turns; mean of three runs at temperature 0.7; higher is better. Source: arxiv.org. Saturation forecast: Around July 2028. 23 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.6 | 41.78 |
| 2 | GPT-5.4 | 39.91 |
| 3 | Gemini 3.1 Pro (Preview) | 37.56 |
| 4 | DeepSeek V3.2 | 36.62 |
| 5 | Claude Sonnet 4.6 | 35.21 |
| 6 | Qwen 3.6 Plus | 35.21 |
| 7 | GLM-5.1 | 35.21 |
| 8 | DeepSeek V4 Pro | 32.86 |
| 9 | GLM-5 | 30.05 |
| 10 | Gemini 2.5 Flash | 28.17 |
| 11 | Qwen 3.5 9B | 27.23 |
| 12 | MiniMax-M2.5 | 26.76 |
| 13 | MiniMax-M2.7 | 26.29 |
| 14 | Kimi K2.5 | 25.35 |
| 15 | Llama 3.3 70B | 22.07 |
Interactive version: theaggregate.ai/benchmark?slug=stt-arena-easy · How It Works · Data refreshed daily, snapshot 2026-10-07.