- MMLU (5-shot) 78.4
- MMLU-Pro (5-shot, CoT) 55.5
- BBH (3-shot, CoT) 81.6
- AGI EVAL (0-shot, CoT) 70.6
- GPQA (0-shot, CoT; subset not stated) 40.6
- AlpacaEval 2.0 42.5
- IFEval Average 87.5
- ArenaHard 46.5
- GSM8K (8-shot) 87.3
- GSM8K Symbolic (8-shot) 87.4
- Minerva Math (0-shot, CoT) 74
- DeepMind Math (0-shot, CoT) 59.3
- HumanEval (pass@1) 88
- HumanEval+ (pass@1) 83
- MBPP (pass@1) 84
- MBPP+ (pass@1) 71
- CRUXEval-O (pass@1) 50.3
- BigCodeBench (pass@1) 46.2
- BFCL v3 64.7
- MULTIPLE (pass@1) 57.4
- MMMLU (5-shot, 11 languages) 69.7
- INCLUDE (5-shot, 14 languages) 64
- MGSM (8-shot, 5 languages) 38.7
- SALAD-Bench 97.3
- AttaQ 86.6
Granite 4.0 H Small
Granite 4.0 H Small places in the top three on AttaQ.
35
Benchmarks with a score
0
Ranked first, out of five or more models
1
In the top three
IBM Oct 2025
Granite 4.0 H Small
Open weights Flagship
No Epoch Capabilities Index score yet
- SimpleQA Verified 8.57% #50 /56
- BigCodeBench 46.2% #5 /11
- MMLU-Pro 55.5% #135 /151
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Vectara Hallucination Leaderboard (2025 dataset) | 5.20% | #13 of 104 | Official leaderboard | |
| AGIEval | 70.6% | #9 of 35 | Model card | |
| IFEval | 87.5% | #51 of 126 | Model card | |
| BIG-Bench Hard | 81.6% | #31 of 71 | Model card | |
| MMLU | 78.4% | #80 of 127 | Model card | |
| LMArena (Elo) | 1,287 | #239 of 333 | Official leaderboard | |
| INCLUDE | 64% | #20 of 25 | Model card | |
| AlpacaEval 2.0 | 42.5% | #13 of 16 | Model card | |
| MMMLU | 69.7% | #53 of 62 | Model card | |
| Wer wird Millionär? LLM Benchmark | 149.0 | #65 of 76 | Official leaderboard | |
| Global MMLU Lite | 75% | #37 of 43 | Official leaderboard | |
| SimpleQA Verified | 8.57% | #50 of 56 | Official leaderboard | |
| MMLU-Pro | 55.5% | #135 of 151 | Model card | |
| Arena-Hard-Auto | 46.5% | #35 of 39 | Model card | |
| MGSM | 38.7% | #49 of 52 | Model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GSM-Symbolic | 87.4% | #1 of 3 | Model card | |
| MATH | 74% | #33 of 93 | Model card | |
| DeepMind Mathematics | 59.3% | #2 of 4 | Model card | |
| GSM8K | 87.3% | #49 of 87 | Model card | |
| GPQA Diamond | 40.6% | #254 of 284 | Model card |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MBPP | 84% | #11 of 66 | Model card | |
| HumanEval | 88% | #27 of 95 | Model card | |
| BigCodeBench | 46.2% | #5 of 11 | Model card | |
| HumanEval+ | 83% | #16 of 27 | Model card | |
| MBPP+ | 71% | #11 of 15 | Model card | |
| CRUXEval | 50.3% | #8 of 9 | Model card | |
| MultiPL-E | 57.4% | #17 of 18 | Model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Berkeley Function Calling Leaderboard (BFCL) | 64.7% | #34 of 47 | Model card |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Anthropic Red Team | 99.8% | #6 of 75 | Official leaderboard | |
| AIR-Bench 2024 | 82% | #20 of 82 | Official leaderboard | |
| SimpleSafetyTests | 100% | #23 of 75 | Official leaderboard | |
| SALAD-Bench | 97.3% | #1 of 3 | Model card | |
| AttaQ | 86.6% | #2 of 5 | Model card | |
| HELM Safety | 91.4 | #38 of 75 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| HELM Capabilities | 57.5 | #46 of 60 | Official leaderboard |
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.