Granite 4.1 3B: benchmark results
IBM's 3B dense decoder-only model from the Apache-2.0 Granite 4.1 family (April 2026), trained on ~15T tokens with context extension up to 512K. Provider: IBM. Released 2026-04-29. Access: Open.
Aggregate Rating 1461 ± 26, rank #929 of 1656 rated models, from 61 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Kurakurai French LLM - MBPP | 68.07 | pass@1 (%) | 95.1 |
| Kurakurai French LLM - IFEval | 75.87 | Accuracy (%) | 93.9 |
| Kurakurai French LLM - BFCL v2 | 72.75 | Accuracy (%) | 93.1 |
| Kurakurai French LLM - HumanEval+ | 68.81 | pass@1 (%) | 92.7 |
| Kurakurai French LLM - Global MMLU | 61.1 | Accuracy (%) | 85.4 |
| Kurakurai French LLM - MGSM | 81.2 | Accuracy (%) | 85.4 |
| Kurakurai French LLM - MMLU-Pro | 43.2 | Accuracy (%) | 85.4 |
| Kurakurai French LLM - GPQA Diamond | 28.6 | Accuracy (%) | 75.6 |
| Kurakurai French LLM - MATH-500 | 55.32 | Accuracy (%) | 70.7 |
| Kurakurai French LLM - AIME 2025 | 2.33 | Accuracy (%) | 63.4 |
| Kurakurai French LLM - MultiIF | 39.4 | Accuracy (%) | 58.5 |
| Kurakurai French LLM - AIME 2024 | 1.67 | Accuracy (%) | 56.1 |
Interactive version: theaggregate.ai/model?slug=granite-4-1-3b · How It Works · Data refreshed daily, snapshot 2026-10-20.