Granite 4.2 30B: benchmark results
IBM's Apache-2.0 30B dense reasoning model from the Granite 4.2 line (August 2026), post-trained from Granite-4.1-30B-Base with a built-in chain-of-thought mode and 128K native context. Provider: IBM. Released 2026-08-25. Access: Open.
Aggregate Rating 1603 ± 27, rank #428 of 1656 rated models, from 92 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| DoubtLess Evaluation - AIME25 and AMC23 - DoubtLess | 91.4 | Accuracy (%) | 100 |
| DoubtLess Evaluation - AIME25 and AMC23 - Plain | 89 | Accuracy (%) | 100 |
| LLM Stats (HMMT25; February 2025 (Granite evaluation)) | 89.17 | Score (%) | 100 |
| AA-Omniscience Hallucination Rate | 25.58 | Hallucination Rate (%) | 95.9 |
| AA-Omniscience Index - Law | -5.7 | Omniscience Index | 75.4 |
| AA-Omniscience Index - Humanities & Social Sciences | -6.7 | Omniscience Index | 70.8 |
| AA-Omniscience Index - Business | -9.8 | Omniscience Index | 68.6 |
| AA-Omniscience Index - Health | -17 | Omniscience Index | 68.3 |
| AA-Omniscience Index - Software Engineering (SWE) - Julia | -24 | Omniscience Index | 63.9 |
| AA Omniscience | -12.92 | Score | 62.9 |
| AA-Omniscience Index - Science, Engineering & Mathematics | -9.7 | Omniscience Index | 62 |
| AA-Omniscience Index - Software Engineering (SWE) - Java | -31 | Omniscience Index | 61.5 |
Interactive version: theaggregate.ai/model?slug=granite-4-2-30b · How It Works · Data refreshed daily, snapshot 2026-10-20.