← All comparisons
Step 3.5 Flash 2603 vs Claude 4.1 Opus (Non-reasoning)
StepFun vs Anthropic — side-by-side benchmark comparison
| Step 3.5 Flash 2603 | Claude 4.1 Opus (Non-reasoning) | |
|---|---|---|
| Intelligence Index | 38.5 | 36.0 |
| Coding Index | 34.6 | — |
| Math Index | — | — |
| Output speed (tok/s) | 167.8 | 44.7 |
| Blended price ($/1M) | $0.00 | $32.81 |
| Time to first token (s) | 0.88s | 1.63s |
| aime | — | — |
| aime 25 | — | — |
| artificial analysis coding index | 34.60 | — |
| artificial analysis intelligence index | 38.50 | 36.00 |
| artificial analysis math index | — | — |
| gpqa | 82.6% | — |
| hle | 22.6% | — |
| ifbench | 66.5% | — |
| lcr | 54.3% | — |
| livecodebench | — | — |
| math 500 | — | — |
| mmlu pro | — | — |
| scicode | 38.5% | — |
| tau2 | 87.4% | — |
| terminalbench hard | 32.6% | — |
Benchmark data from Artificial Analysis.