CARTE-LV: leaderboard
Metric: Accuracy (%) on the 233 CARTE-LV questions on regional linguistic variation (grammar, discourse markers, pragmatics, register and region-specific lexical usage in 13 regions), 0-shot, five options including 'Je ne sais pas'; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 27 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash | 93.6 |
| 2 | Mistral Nemo Instruct (2407) | 74.2 |
| 3 | Gemma 3 12B (IT) | 68.7 |
| 4 | Qwen 3.5 9B | 65.7 |
| 5 | Llama 3.1 8B Instruct | 63.9 |
| 6 | Qwen 3.5 4B | 58.4 |
| 7 | aya-expanse-8B | 58.4 |
| 8 | Mistral 7B Instruct (v0.2) | 57.9 |
| 9 | Mistral 7B Instruct (v0.3) | 57.5 |
| 10 | Luth-LFM2-1.2B | 49.4 |
| 11 | Llama 3.2 3B Instruct | 48.9 |
| 12 | Lucie-7B | 44.2 |
| 13 | Llama 3.2 1B Instruct | 40.8 |
| 14 | gemma-4-E4B-it | 23.2 |
| 15 | Mistral-7B-v0.1 | 23.2 |
Interactive version: theaggregate.ai/benchmark?slug=carte-lv · How It Works · Data refreshed daily, snapshot 2026-09-29.