← All comparisons
Granite 4.1 3B vs Claude 3.5 Haiku
IBM vs Anthropic — side-by-side benchmark comparison
| Granite 4.1 3B | Claude 3.5 Haiku | |
|---|---|---|
| Intelligence Index | 8.5 | 18.7 |
| Coding Index | 5.5 | 10.7 |
| Math Index | — | — |
| Output speed (tok/s) | 0.0 | 0.0 |
| Blended price ($/1M) | $0.00 | $1.75 |
| Time to first token (s) | 0.00s | 0.00s |
| aime | — | 3.3% |
| aime 25 | — | — |
| artificial analysis coding index | 5.50 | 10.70 |
| artificial analysis intelligence index | 8.50 | 18.70 |
| artificial analysis math index | — | — |
| gpqa | 31.4% | 40.8% |
| hle | 3.4% | 3.5% |
| ifbench | 33.7% | 42.8% |
| lcr | 3.0% | 23.3% |
| livecodebench | — | 31.4% |
| math 500 | — | 72.1% |
| mmlu pro | — | 63.4% |
| scicode | 11.9% | 27.4% |
| tau2 | 19.6% | 24.6% |
| terminalbench hard | 2.3% | 2.3% |
Benchmark data from Artificial Analysis.