benchmarks.site

Public leaderboard · Multilingual

EuroEval

Tests how well AI language models understand and write in more than 30 European languages, across many kinds of tasks. It matters because most AI tests are English-only.

Leaderboard

Top 3 right now

  1. 1stgpt-6-astra (zero-shot, val)1.17 ± 0.05
  2. 2ndgpt-6.1-sol (zero-shot, val)1.18 ± 0.04
  3. 3rdgemini/gemini-3.7-flash (zero-shot, val)1.24 ± 0.03

What it measures

Language models on tasks such as classification, NER, reading comprehension and knowledge across European languages, with per-language leaderboards.

How to read the numbers

Average rank score across languages and tasks (lower is better; 1.00 would mean best on everything) — lower is better.

Results

25 rows · lower is better (Mean rank score)CSV
1gpt-6-astra (zero-shot, val)1.17 ± 0.05No
1gpt-6.1-sol (zero-shot, val)1.18 ± 0.04No
2gemini/gemini-3.7-flash (zero-shot, val)1.24 ± 0.03No
3gemini/gemini-3-flash-preview#thinking (zero-shot, val)1.32 ± 0.04No
3gemini/gemini-3.6-flash (zero-shot, val)1.31 ± 0.04No
3zai-org/GLM-5.3-Flash (val)1.36 ± 0.04Yes
4gpt-5.6-sol (zero-shot, val)1.40 ± 0.04No
5gpt-6-sol (zero-shot, val)1.45 ± 0.05No
6zai-org/GLM-5.3-Flash (zero-shot, val)1.54 ± 0.08Yes
6deepseek/deepseek-flash#low (zero-shot, val)1.50 ± 0.05No
7Qwen/Qwen3.6-27B-FP8 (val)1.60 ± 0.07Yes
7gemini/gemini-3.1-flash-lite-preview (zero-shot, val)1.59 ± 0.08No
7gpt-5.6-terra (zero-shot, val)1.58 ± 0.07No
7openai/gpt-5.4-mini-2026-03-17#high (zero-shot, val)1.61 ± 0.05No
7claude-sonnet-4-5-20250929#thinking (zero-shot, val)1.56 ± 0.08No
7deepseek/deepseek-flash#no-thinking (val)1.57 ± 0.06No
8Qwen/Qwen3.8-27B (val)1.67 ± 0.09Yes
8Qwen/Qwen3.5-35B-A3B-FP8 (val)1.66 ± 0.09Yes
8google/gemma-4-31B-it (val)1.67 ± 0.10Yes
8claude-sonnet-4-6 (zero-shot, val)1.69 ± 0.08No
8xai/grok-4-1-fast-reasoning (zero-shot, val)1.69 ± 0.07No
8openai/gpt-5.4-mini-2026-03-17#medium (zero-shot, val)1.64 ± 0.06No
9gpt-5-mini-2025-08-07 (zero-shot, val)1.71 ± 0.07No
9openai/gpt-5.4-mini-2026-03-17#low (zero-shot, val)1.70 ± 0.06No
9openai/gpt-5.4-2026-03-05#none (zero-shot, val)1.76 ± 0.09No

Note: European (all languages) leaderboard, All Models view, simplified CSV embedded in the site bundle. Rank score: lower is better. Ties share ranks.

Source: EuroEval team · euroeval.com · results as of 2026-11-15 · retrieved 2026-11-17

Markdown · More Multilingual