Polish Multi-Turn Benchmark.

Polish adaptation of MT-Bench evaluating LLMs on multi-turn conversation quality across 8 categories: coding, extraction, humanities, math, reasoning, roleplay, STEM, and writing. Scores on a 1-10 scale judged by GPT-4. Created by SpeakLeash.

Best published scores.

450 results indexed across 9 metrics. Shaded row marks current SOTA; ties broken by submission date.

Primary

pl-score · higher is better

All metrics

coding, extraction, humanities, math, pl-score, reasoning, roleplay, stem, writing

1 steps
of state of the art.

Each row below marks a model that broke the previous record on pl-score. Intermediate submissions are kept in the leaderboard above; only SOTA-setting entries are re-listed here.

Higher scores win. Each subsequent entry improved upon the previous best.

1. Apr 2, 2026Gemma 3 (27B, IT)Google9.28

Have a score that beats
this table?

Submit a checkpoint and a reproduction script. We will run it, publish the score, and — if it takes the top — annotate the step on the progress chart with your name.

• 01A public checkpoint or API endpoint
• 02A reproduction script with frozen commit + seed
• 03Declared evaluation environment (Python, deps)
• 04One row per metric declared by this dataset
• 05A contact so we can follow up on discrepancies