Granite 4.0 350M: benchmark results
IBM's Granite 4.0 350M instruct model, fine-tuned from Granite-4.0-350M-Base with supervised fine-tuning, reinforcement learning and model merging, part of the Granite 4.0 Nano collection. Provider: IBM. Released 2025-10-28. Access: Open.
Aggregate Rating 1339 ± 38, rank #1410 of 1656 rated models, from 55 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA-Omniscience Hallucination Rate | 76.13 | Hallucination Rate (%) | 55.2 |
| AA Humanity's Last Exam | 5.51 | Accuracy (%) | 32.9 |
| AA CritPt | 0 | Accuracy (%) | 22 |
| AA Omniscience - Software Engineering (SWE) - Go | 8 | Accuracy (%) | 20 |
| AA-Omniscience Index - Health | -63.6 | Omniscience Index | 15.7 |
| AA-Omniscience Index - Business | -61.9 | Omniscience Index | 14.9 |
| CritPt | 0 | Accuracy (imported) | 14.2 |
| AA Omniscience - Software Engineering (SWE) - R | 2 | Accuracy (%) | 14 |
| AA Omniscience - Software Engineering (SWE) - Kotlin | 8 | Accuracy (%) | 13.7 |
| AA Omniscience - Software Engineering (SWE) - Swift | 16 | Accuracy (%) | 13.5 |
| AA-Omniscience Index - Software Engineering (SWE) - Julia | -88 | Omniscience Index | 13 |
| AA-Omniscience Index - Software Engineering (SWE) - Kotlin | -70 | Omniscience Index | 12.4 |
Interactive version: theaggregate.ai/model?slug=granite-4-0-350m · How It Works · Data refreshed daily, snapshot 2026-10-20.