← All comparisons
Phi-4 Multimodal Instruct vs DeepSeek R1 0528 Qwen3 8B
Microsoft vs DeepSeek — side-by-side benchmark comparison
| Phi-4 Multimodal Instruct | DeepSeek R1 0528 Qwen3 8B | |
|---|---|---|
| Intelligence Index | 10.0 | 16.4 |
| Coding Index | — | 7.8 |
| Math Index | — | 63.7 |
| Output speed (tok/s) | 16.6 | 0.0 |
| Blended price ($/1M) | $0.00 | $0.00 |
| Time to first token (s) | 1.33s | 0.00s |
| aime | 9.3% | 65.0% |
| aime 25 | — | 63.7% |
| artificial analysis coding index | — | 7.80 |
| artificial analysis intelligence index | 10.00 | 16.40 |
| artificial analysis math index | — | 63.70 |
| gpqa | 31.5% | 61.2% |
| hle | 4.4% | 5.6% |
| ifbench | — | 19.9% |
| lcr | — | 13.0% |
| livecodebench | 13.1% | 51.3% |
| math 500 | 69.3% | 93.2% |
| mmlu pro | 48.5% | 73.9% |
| scicode | 11.0% | 20.4% |
| tau2 | — | 0.0% |
| terminalbench hard | — | 1.5% |
Benchmark data from Artificial Analysis.