MARS-Bench - Task Switching: leaderboard
Each model, in an ultra multi-turn chat over the text play-by-play of 30 games (15 English NBA, 10 Chinese NBA, 5 English NHL), must answer the match questions interleaved with unrelated math questions (the score covers the match questions only; the paper reports the math separately); the score is the category score. A DeepSeek-V3-0324 judge awards checklist points built from the play-by-play.
Metric: Score (0-100). Source: arxiv.org. Saturation forecast: Estimated already saturated. 13 models tracked. Updated 17 Nov 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro (Preview 05-06) | 66.72 | #271 |
| 2 | O1 (2024-12-17) | 58.28 | #314 |
| 3 | Doubao-1.5-Thinking-Pro | 50.69 | #257 |
| 4 | GPT-4.5 (Preview) | 40.74 | #499 |
| 5 | DeepSeek R1 | 39.4 | #487 |
| 6 | GPT-4o (2024-11-20) | 36.12 | #864 |
| 7 | Claude 3.7 Sonnet | 34.15 | #476 |
| 8 | Doubao-1.5-Pro-32k | 32.28 | #696 |
| 9 | DeepSeek V3 (1226) | 29.63 | #807 |
| 10 | GPT-4.1 Mini | 28.13 | #793 |
| 11 | Qwen 2.5 72B Instruct | 23.97 | #1119 |
| 12 | Qwen 2.5 Max | 23.22 | #966 |
| 13 | GLM-Z1-Air | 20.76 |
No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=mars-bench-task-switching · How It Works · Data refreshed daily, snapshot 2026-11-17.