granite-4.1-3B-base: benchmark results

Provider: IBM. Access: Open.

Aggregate Rating 1403 ± 37, rank #6109 of 10552 rated models, from 26 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Lara Mini Base Report (base-model protocol) - CRUXEval-I (2-shot)44.9pass@1, direct prediction, official answer parser (%)92.6
Lara Mini Base Report (base-model protocol) - CRUXEval-O (2-shot)42.5pass@1, direct prediction, official answer parser (%)90.7
Lara Mini Base Report (base-model protocol) - GSM8K (5-shot)72.2Exact match (strict extraction: first "#### N", else first \83.3
Lara Mini Base Report (base-model protocol) - MGSM (8 (CoT)-shot)47.3Exact match on each language's native CoT answer phrase, ave81.5
Lara Mini Base Report (base-model protocol) - PIQA (0-shot)79.5Accuracy normalized by answer length (%), English benchmark79.6
Lara Mini Base Report (base-model protocol) - MBPP (3-shot)51.4pass@1 (execution) (%)77.8
Lara Mini Base Report (base-model protocol) - ARC-Challenge (25-shot)59.9Accuracy normalized by answer length (%), English benchmark75.9
Lara Mini Base Report (base-model protocol) - MMLU (5-shot)64.3Accuracy (%), English benchmark75.9
Lara Mini Base Report (base-model protocol) - WinoGrande (5-shot)72.9Accuracy (%), English benchmark74.1
Lara Mini Base Report (base-model protocol) - MMLU-Pro (5 (CoT)-shot)37.4Exact match (%), English benchmark73.1
Lara Mini Base Report (base-model protocol) - Global MMLU, All Languages (5-shot)49.9Accuracy (%) averaged over all languages/variants of the ben72.2
Lara Mini Base Report (base-model protocol) - HellaSwag (10-shot)75.7Accuracy normalized by answer length (%), English benchmark71.3

Interactive version: theaggregate.ai/model?slug=granite-4-1-3b-base · How It Works · Data refreshed daily, snapshot 2026-11-16.