Vector Eval - MMLU: leaderboard
Vector Institute Inspect Evals run of MMLU, testing multitask academic knowledge across humanities, STEM, social science, and professional subjects.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet | 86.65 | #350 |
| 2 | Gemini 1.5 Pro | 84.67 | #445 |
| 3 | GPT-4o | 84.35 | #343 |
| 4 | Mistral Large 2 (Jul) | 80.36 | #484 |
| 5 | Llama 3.1 70B Instruct | 80.34 | #561 |
| 6 | Gemini 1.5 Flash | 77.15 | #618 |
| 7 | GPT-4o Mini | 76.98 | #599 |
| 8 | Command-R+ | 69.51 | #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 · How It Works · Data refreshed daily, snapshot 2026-10-20.