MT-Bench PL - STEM: leaderboard
The "STEM" subtask of MT-Bench PL. The Polish-language version of MT-Bench, scoring multi-turn conversational and instruction-following ability via LLM-as-judge.
Metric: Judge Score (0-10). Source: huggingface.co. Saturation forecast: Estimated already saturated. 50 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 3 12B (IT) | 10 | #664 |
| 2 | Phi-4 | 10 | #706 |
| 3 | Gemma 3 27B (IT) | 9.95 | #523 |
| 4 | aya-expanse-32B | 9.95 | #692 |
| 5 | Mistral Small 3.1 | 9.9 | #610 |
| 6 | Gemma 2 27B (IT) | 9.8 | #645 |
| 7 | aya-expanse-8B | 9.75 | #931 |
| 8 | Qwen 2.5 32B Instruct | 9.7 | #498 |
| 9 | Gemma 3 4B (IT) | 9.65 | #992 |
| 10 | Mistral-Small-Instruct-2409 | 9.65 | #702 |
| 11 | Qwen 2.5 14B Instruct | 9.6 | #647 |
| 12 | Llama 3.1 70B Instruct | 9.55 | #561 |
| 13 | Qwen 2 72B Instruct | 9.55 | #576 |
| 14 | Mistral Small 3 | 9.5 | #627 |
| 15 | Bielik-11B-v2.2-Instruct | 9.45 | #660 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mt-bench-pl-stem · How It Works · Data refreshed daily, snapshot 2026-10-20.