granite-3.0-3B-a800m-instruct: benchmark results
IBM's Granite 3.0 3B-A800M Instruct, a mixture-of-experts model (3B total, about 800M active) fine-tuned from its 4K-context base with supervised fine-tuning, reinforcement learning and model merging. Provider: IBM. Released 2024-10-21. Access: Open.
Aggregate Rating 1334 ± 28, rank #1418 of 1656 rated models, from 54 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Enkrypt AI - Toxicity Risk | 0.09 | Risk Score | 96.1 |
| Enkrypt AI - Safety Risk | 22.78 | Risk Score | 70.7 |
| EuroEval Spanish NLU - MLQA ES | 62.81 | Character F1 (%) | 68.1 |
| EuroEval Portuguese NLU - SST-2 PT | 79.46 | Matthews correlation coefficient (%) | 62.7 |
| EuroEval Dutch NLU - DBRD | 88.43 | Matthews correlation coefficient (%) | 58.7 |
| Enkrypt AI - Jailbreak Risk | 8.88 | Risk Score | 52.9 |
| EuroEval Italian NLU - SQuAD IT | 65.67 | Character F1 (%) | 49.4 |
| Open LLM Leaderboard - IFEval | 42.98 | Score | 44.9 |
| Enkrypt AI - Insecure Code Risk | 34.22 | Risk Score | 41.2 |
| EuroEval Spanish NLU - ScaLA ES | 4.87 | Matthews correlation coefficient (%) | 39.3 |
| EuroEval Italian NLU - Sentipolc16 | 44.07 | Matthews correlation coefficient (%) | 37.6 |
| Open LLM Leaderboard - MATH Level 5 | 7.02 | Score | 37.5 |
Interactive version: theaggregate.ai/model?slug=granite-3-0-3b-a800m-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.