← All comparisons
Gemini 3.1 Flash-Lite vs DeepSeek V4 Pro (Non-reasoning)
Google vs DeepSeek — side-by-side benchmark comparison
| Gemini 3.1 Flash-Lite | DeepSeek V4 Pro (Non-reasoning) | |
|---|---|---|
| Intelligence Index | 33.5 | 39.3 |
| Coding Index | 30.1 | 38.4 |
| Math Index | — | — |
| Output speed (tok/s) | 304.9 | 53.3 |
| Blended price ($/1M) | $0.56 | $0.54 |
| Time to first token (s) | 5.00s | 1.18s |
| aime | — | — |
| aime 25 | — | — |
| artificial analysis coding index | 30.10 | 38.40 |
| artificial analysis intelligence index | 33.50 | 39.30 |
| artificial analysis math index | — | — |
| gpqa | 82.2% | 71.7% |
| hle | 16.2% | 7.7% |
| ifbench | 77.2% | 45.8% |
| lcr | 65.3% | 45.0% |
| livecodebench | — | — |
| math 500 | — | — |
| mmlu pro | — | — |
| scicode | 41.9% | 42.4% |
| tau2 | 31.3% | 91.2% |
| terminalbench hard | 24.2% | 36.4% |
Benchmark data from Artificial Analysis.