EuroEval Estonian NLU - Grammar ET: leaderboard

The "Grammar ET" subtask of EuroEval Estonian NLU. Estonian-language EuroEval natural-language-understanding suite, separating NLU task performance from the broader all-task EuroEval aggregate.

Metric: Matthews correlation coefficient (%). Source: euroeval.com. Saturation forecast: Around August 2027. 250 models tracked.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash (Preview) (Non-reasoning)66.32#82 (Gemini 3 Flash (Preview))
2Gemini 3 Flash (Preview) (Thinking)64.83#82 (Gemini 3 Flash (Preview))
3Gemini 3 Pro (Preview)62.63#66
4Gemini 2.5 Pro61.06#152
5Gemini 3.1 Flash Lite (Preview)58.7#194
6Gemini 2.5 Flash (Thinking)56#245 (Gemini 2.5 Flash)
7GPT-555.2#98
8GPT-5.4 Mini (2026-03-17) (High)53.72#169 (GPT-5.4 Mini (2026-03-17))
9Claude Sonnet 4.652.96#90
10GPT-4.152.89#248
11EuroLLM-22B-Instruct-251251.64#701
12GPT-5.246.89#110
13GPT-5 Mini (2025-08-07)46.35#172
14Gemini 2.5 Flash Lite (Thinking)46.21#422 (Gemini 2.5 Flash Lite)
15Gemma 3 27B (IT)44.41#523

No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.

Interactive version: theaggregate.ai/benchmark?slug=euroeval-estonian-nlu-grammar-et · How It Works · Data refreshed daily, snapshot 2026-10-20.