Vector Eval - MMLU-Pro: leaderboard
Vector Institute Inspect Evals run of MMLU-Pro, using harder multi-choice academic questions with more answer options and stronger reasoning demands.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around January 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O1 | 84.47 | #183 |
| 2 | DeepSeek R1 | 83.83 | #251 |
| 3 | Qwen3 Coder Next | 82.46 | #326 |
| 4 | O3 Mini | 79.25 | #272 |
| 5 | Claude 3.5 Sonnet | 77.63 | #350 |
| 6 | Gemini 1.5 Pro | 75.64 | #445 |
| 7 | GPT-4o | 74.51 | #343 |
| 8 | Mistral Large 2 (Jul) | 69.42 | #484 |
| 9 | Llama 3.1 70B Instruct | 66.89 | #561 |
| 10 | GPT-4o Mini | 63.96 | #599 |
| 11 | Gemini 1.5 Flash | 59.93 | #618 |
| 12 | Command-R+ | 44.12 | #891 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=vector-eval-mmlu-pro · How It Works · Data refreshed daily, snapshot 2026-10-20.