LFM2.5-8B-A1B-Base: benchmark results

Provider: Liquid AI. Access: Open.

Aggregate Rating 1399 ± 36, rank #6243 of 10552 rated models, from 26 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Lara Mini Base Report (base-model protocol) - ARC-Challenge (25-shot)64.9Accuracy normalized by answer length (%), English benchmark94.4
Lara Mini Base Report (base-model protocol) - MMLU-Pro (5 (CoT)-shot)39.8Exact match (%), English benchmark87
Lara Mini Base Report (base-model protocol) - MATH (4-shot)45.8Math-Verify equivalence, Minerva prompts, 5,000 problems ove79.6
Lara Mini Base Report (base-model protocol) - Global MMLU, All Languages (5-shot)50.3Accuracy (%) averaged over all languages/variants of the ben77.8
Lara Mini Base Report (base-model protocol) - GSM8K (5-shot)67.6Exact match (strict extraction: first "#### N", else first \75.9
Lara Mini Base Report (base-model protocol) - COPA (0-shot)85.4Accuracy (%), English benchmark73.1
Lara Mini Base Report (base-model protocol) - MGSM (8 (CoT)-shot)43.3Exact match on each language's native CoT answer phrase, ave70.4
Lara Mini Base Report (base-model protocol) - MMLU (5-shot)63.2Accuracy (%), English benchmark70.4
Lara Mini Base Report (base-model protocol) - XCOPA, All Languages (0-shot)58.6Accuracy (%) averaged over all languages/variants of the ben69.4
Lara Mini Base Report (base-model protocol) - Global PIQA (non-parallel), All Languages (0-shot)56.9Accuracy (%) averaged over all languages/variants of the ben68.5
Lara Mini Base Report (base-model protocol) - Global PIQA (parallel), All Languages (0-shot)25.1Accuracy (%) averaged over all languages/variants of the ben66.7
Lara Mini Base Report (base-model protocol) - INCLUDE, All Languages (5-shot)44.5Accuracy (%) averaged over all languages/variants of the ben66.7

Interactive version: theaggregate.ai/model?slug=lfm2-5-8b-a1b-base · How It Works · Data refreshed daily, snapshot 2026-11-16.