← All benchmarks
DeepSeek V4 Pro (Reasoning, Max Effort)
DeepSeek · Released 2026-04-24 · Independent benchmark scores from Artificial Analysis
Intelligence Index
44.3
Output speed
71.8
tokens/sec
Blended price
$0.54
per 1M tokens
Time to first token
0.91s
Model metadata
| Model ID | bf220674-68bd-43cc-a1b8-ce5ed4d2f18d |
|---|---|
| Slug | deepseek-v4-pro |
| Release date | 2026-04-24 |
| Provider | DeepSeek |
| Provider slug | deepseek |
| Output speed | 71.8 tok/s |
| Time to first token | 0.91s |
| Time to first answer | 61.83s |
| Blended price ($/1M) | $0.54 |
| Input price ($/1M) | $0.43 |
| Output price ($/1M) | $0.87 |
All evaluation scores (17)
| aime | — |
|---|---|
| aime 25 | — |
| artificial analysis coding index | 59.40 |
| artificial analysis intelligence index | 44.30 |
| artificial analysis math index | — |
| gpqa | 88.8% |
| hle | 35.9% |
| ifbench | 76.5% |
| lcr | 66.3% |
| livecodebench | — |
| math 500 | — |
| mmlu pro | — |
| scicode | 50.0% |
| tau banking | 25.8% |
| tau2 | 96.2% |
| terminalbench hard | 46.2% |
| terminalbench v2 1 | 64.0% |
Source: artificialanalysis.ai/models/deepseek-v4-pro
Benchmark data from Artificial Analysis.