LFM2 2.6B: benchmark results
Liquid AI's LFM2-2.6B, the largest of four LFM2 checkpoints from 350M to 2.6B parameters, built with a hybrid architecture of multiplicative gates and short convolutions for edge deployment. Provider: Liquid AI. Released 2025-09-23. Access: Open.
Aggregate Rating 1369 ± 14, rank #1336 of 1656 rated models, from 412 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Kurakurai French LLM - Global MMLU | 65.9 | Accuracy (%) | 92.7 |
| Kurakurai French LLM - AIME 2024 | 14 | Accuracy (%) | 90.2 |
| Kurakurai French LLM - AIME 2025 | 11 | Accuracy (%) | 90.2 |
| Kurakurai French LLM - MATH-500 | 71.62 | Accuracy (%) | 90.2 |
| Kurakurai French LLM - GPQA Diamond | 30.8 | Accuracy (%) | 87.8 |
| Kurakurai French LLM - MGSM | 83.6 | Accuracy (%) | 87.8 |
| Kurakurai French LLM - MMLU-Pro | 45.1 | Accuracy (%) | 87.8 |
| Kurakurai French LLM - MultiIF | 55.85 | Accuracy (%) | 82.9 |
| EuroEval Spanish NLU - Sentiment Headlines ES | 46.63 | Matthews correlation coefficient (%) | 75.7 |
| Kurakurai French LLM - IFEval | 68.26 | Accuracy (%) | 75.6 |
| Kurakurai French LLM - HumanEval+ | 52.12 | pass@1 (%) | 73.2 |
| Kurakurai French LLM - MBPP | 54.42 | pass@1 (%) | 73.2 |
Interactive version: theaggregate.ai/model?slug=lfm2-2-6b · How It Works · Data refreshed daily, snapshot 2026-10-20.