← All comparisons
Ministral 3 3B vs Llama 3.1 Nemotron Ultra 253B v1 (Reasoning)
Mistral vs NVIDIA — side-by-side benchmark comparison
| Ministral 3 3B | Llama 3.1 Nemotron Ultra 253B v1 (Reasoning) | |
|---|---|---|
| Intelligence Index | 11.2 | 15.0 |
| Coding Index | 4.8 | 13.1 |
| Math Index | 22.0 | 63.7 |
| Output speed (tok/s) | 200.8 | 52.3 |
| Blended price ($/1M) | $0.10 | $0.90 |
| Time to first token (s) | 0.34s | 0.76s |
| aime | — | 74.7% |
| aime 25 | 22.0% | 63.7% |
| artificial analysis coding index | 4.80 | 13.10 |
| artificial analysis intelligence index | 11.20 | 15.00 |
| artificial analysis math index | 22.00 | 63.70 |
| gpqa | 35.8% | 72.8% |
| hle | 5.3% | 8.1% |
| ifbench | 26.8% | 38.2% |
| lcr | 11.7% | 7.3% |
| livecodebench | 24.7% | 64.1% |
| math 500 | — | 95.2% |
| mmlu pro | 52.4% | 82.5% |
| scicode | 14.4% | 34.7% |
| tau2 | 24.9% | 11.4% |
| terminalbench hard | 0.0% | 2.3% |
Benchmark data from Artificial Analysis.