BlueBench: leaderboard
IBM's BlueBench — broad enterprise-oriented evaluation suite (run via unitxt) covering reasoning, knowledge, safety and domain tasks.
Metric: Average Score (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 | 64.61 | #248 |
| 2 | GPT-4o | 64.13 | #343 |
| 3 | GPT-4.1 Mini | 63.01 | #355 |
| 4 | Mistral Medium 3 | 60.5 | #500 |
| 5 | Llama 3.3 70B Instruct | 58.96 | #534 |
| 6 | O1 | 58.6 | #183 |
| 7 | GPT-4.1 Nano | 56.98 | #731 |
| 8 | Mistral Large | 55.49 | #710 |
| 9 | O4 Mini | 53.98 | #180 |
| 10 | O3 Mini | 53.85 | #272 |
| 11 | Granite 3.3 8B Instruct | 49.29 | #825 |
| 12 | Pixtral-12B | 45.38 | #808 |
| 13 | Llama 3.2 3B Instruct | 44.27 | #1352 |
| 14 | Llama 3.2 1B Instruct | 31.83 | #1539 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=bluebench · How It Works · Data refreshed daily, snapshot 2026-10-20.