Across 10 shared benchmarks, Claude Sonnet 4 scores higher on 7 and Granite 4.2 8B on 3. The widest gap is τ³-Bench, where Granite 4.2 8B scores 66.3 against 13.8.
| Benchmark | Claude Sonnet 4 | Granite 4.2 8B |
|---|---|---|
| gdpval | 31.2 | 1189 |
| GPQA Diamond | 78 | 64.1 |
| LiveCodeBench v6 | 48.5 | 73.2 |
| MMLU-Pro | 84 | 74 |
| scicode | 40 | 36.1 |
| SWE-bench Multilingual | 56.9 | 30.8 |
| SWE-bench Pro | 42.7 | 19.1 |
| SWE-bench Verified | 80.2 | 47.7 |
| Terminal-Bench 2.1 | 36.3 | 20.6 |
| τ³-Bench | 13.8 | 66.3 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.