Magis-Bench — leaderboard
Metric: AVG (self-reported). Source: benchmarklist.com. 23 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 6.67 |
| 2 | Claude Opus 4.5 | 6.46 |
| 3 | GPT-5.1 | 6.23 |
| 4 | Gemini 2.5 Pro | 6.18 |
| 5 | Claude Sonnet 4.5 | 5.55 |
| 6 | GPT-4.1 | 5.2 |
| 7 | DeepSeek V3.2 | 4.34 |
| 8 | Kimi K2 (Thinking) | 4.14 |
| 9 | GPT-5 Mini | 4.08 |
| 10 | GPT-4.1 Mini | 3.84 |
| 11 | Qwen 2.5 72B Instruct | 2.62 |
| 12 | Qwen 2.5 14B Instruct | 2.04 |
| 13 | Qwen 3 8B | 1.82 |
Interactive version: theaggregate.ai/benchmark?slug=magis-bench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.