EuroEval Latvian NLU - ScaLA LV: leaderboard
The "ScaLA LV" subtask of EuroEval Latvian NLU. Latvian-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 March 2028. 233 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 58.1 | #66 |
| 2 | GPT-5.2 | 56.85 | #110 |
| 3 | Gemini 2.5 Flash (Thinking) | 56.6 | #245 (Gemini 2.5 Flash) |
| 4 | Gemini 3 Flash (Preview) (Thinking) | 55.05 | #82 (Gemini 3 Flash (Preview)) |
| 5 | Gemini 3.1 Flash Lite (Preview) | 52.6 | #194 |
| 6 | GPT-5 | 50.23 | #98 |
| 7 | Claude Sonnet 4.6 | 49.49 | #90 |
| 8 | Gemini 2.5 Pro | 49.04 | #152 |
| 9 | Gemini 3 Flash (Preview) (Non-reasoning) | 47.78 | #82 (Gemini 3 Flash (Preview)) |
| 10 | GPT-4.1 | 46.08 | #248 |
| 11 | Qwen 3 235B A22B 2507 Instruct | 44.64 | #300 |
| 12 | Qwen 3 32B (Non-reasoning) | 43.82 | #443 (Qwen 3 32B) |
| 13 | GPT-5.4 Nano (2026-03-17) (High) | 42.54 | #266 (GPT-5.4 Nano (2026-03-17)) |
| 14 | GPT-5 Mini (2025-08-07) | 41.23 | #172 |
| 15 | Claude Sonnet 4.5 (Thinking) | 40.63 | #142 (Claude Sonnet 4.5) |
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-nlu-scala-lv · How It Works · Data refreshed daily, snapshot 2026-10-20.