Granite 4.0H Tiny: benchmark results
Provider: IBM. Released 2025-10-02. Access: Open.
Aggregate Rating 1401 ± 28, rank #6075 of 10916 rated models, from 34 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Medmarks - MedHallu Hard | 44.95 | Score (%) | 48.9 |
| Medmarks - MedHallu Medium | 51.67 | Score (%) | 28.4 |
| Medmarks - PubMedQA | 69.53 | Score (%) | 22.7 |
| Medmarks - MedHallu Easy | 52.17 | Score (%) | 12.5 |
| Medmarks - MedConceptsQA Easy | 43.4 | Score (%) | 11.4 |
| Medmarks - HEAD-QA v2 | 63.81 | Score (%) | 9.1 |
| Medmarks - Med-HALT Reasoning FCT | 12.96 | Score (%) | 9.1 |
| Medmarks - PubHealthBench Reviewed | 75.92 | Score (%) | 9.1 |
| MTRAG-UN (RAG) | 31 | algorithmic reference-based score RB_alg (MTRAG's combinatio | 8.3 |
| MTRAG-UN (Reference Passages) | 35 | algorithmic reference-based score RB_alg (MTRAG's combinatio | 8.3 |
| MTRAG-UN (Reference Passages) - Faithfulness | 48 | RAGAS faithfulness RL_F (whether the answer is supported by | 8.3 |
| Medmarks - MedConceptsQA Medium | 30.13 | Score (%) | 8 |
Interactive version: theaggregate.ai/model?slug=granite-4-0h-tiny · How It Works · Data refreshed daily, snapshot 2026-11-08.