EuroEval Latvian Common Sense Reasoning: leaderboard
EuroEval's Latvian common-sense reasoning task — multiple-choice questions testing everyday physical and social reasoning in Latvian, from the EuroEval European language model benchmark suite.
Metric: Common Sense Reasoning Average Score (%). Source: euroeval.com. Saturation forecast: Around April 2027. 212 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Mini (2025-08-07) | 97.25 | #172 |
| 2 | Claude Sonnet 4.6 | 94.71 | #90 |
| 3 | Gemini 3 Pro (Preview) | 94.71 | #66 |
| 4 | Gemini 2.5 Pro | 94.34 | #152 |
| 5 | Gemini 3.1 Flash Lite (Preview) | 92.24 | #194 |
| 6 | Gemma 3 27B (IT) | 92.21 | #523 |
| 7 | Gemini 3 Flash (Preview) (Non-reasoning) | 91.68 | #82 (Gemini 3 Flash (Preview)) |
| 8 | GPT-5.4 Mini (2026-03-17) (High) | 91.59 | #169 (GPT-5.4 Mini (2026-03-17)) |
| 9 | GPT-5 | 91.51 | #98 |
| 10 | GPT-5.2 | 91.23 | #110 |
| 11 | gemma-3-27B-pt | 91.19 | #502 |
| 12 | Claude Sonnet 4.5 (Thinking) | 90.43 | #142 (Claude Sonnet 4.5) |
| 13 | Qwen 3 235B A22B 2507 FP8 (Thinking) | 90.41 | #246 (Qwen 3 235B A22B 2507 FP8) |
| 14 | Gemini 3 Flash (Preview) (Thinking) | 90.37 | #82 (Gemini 3 Flash (Preview)) |
| 15 | Qwen 3 235B A22B 2507 Instruct | 89.67 | #300 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=euroeval-latvian-common-sense-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-20.