Head to head
DeepSeek V4-Pro 0813 leads on 15 of 15 shared benchmarks.
DeepSeek V4-Pro 0813 leads on 15 of 15 shared benchmarks.
| Dimension | DeepSeek V4-Pro 0813 | DeepSeek V4-Flash 0731 |
|---|---|---|
| GPQA Diamond | 92.4% | 89.9% |
| HLE | 42.7% | 37.8% |
| Codeforces (Rating) | 3348 | 3289 |
| MathArena Apex | 65.3% | 58.6% |
| Terminal-Bench 2.1 | 87.9% | 82.7% |
| Terminal-Bench 3.0 | 11.8% | 7.6% |
| Terminal-Bench 4.0 | 12.4% | 7% |
| DeepSWE v1.1 | 62.7% | 54.4% |
| ProgramBench | 15.5% | — |
| NL2Repo-Bench | 61.5% | 54.2% |
| CyberGym | 83.3% | 76.7% |
| SEC-Bench Pro | 56.4% | 30.9% |
| ExploitGym | 5.4% | 1.8% |
| HLE (w/tools) | 60% | 51.5% |
| Automation-Bench | 43.2% | 37.7% |
| Agents' Last Exam | 25.7% | 25.2% |
| Blended price | $2.64 /1M | — |
| Full priceInput · Output · Cached input | $1.32 · $3.96 · $0.044 | — |
| Context window | 1,000K | — |
Adjust the monthly volume — every model recomputes instantly.
Estimates use list prices and ignore cached input, batch discounts, and off-peak rates.