DeepSeek-V4-Pro
- Made by
- DeepSeek
- Released
- Announced in
- DeepSeek V4 Preview Releaseapi-docs.deepseek.com
- Event
- DeepSeek V4 Preview
- API record
- /api/v1/models/deepseek-v4-pro
Benchmark results
| Benchmark | Score | Source check |
|---|---|---|
| Humanity's Last ExamNo tools · max reasoning effort; full or text-only set not stated | 37.7% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| Humanity's Last ExamWith tools · max reasoning effort | 48.2% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| SWE-bench Verifiedmax reasoning effort | 80.6% | CheckedA person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| SWE-bench Promax reasoning effort · The source doesn't say whether this is the public or the commercial set | 55.4% | CheckedA person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| GPQA Diamondmax reasoning effort | 90.1% | CheckedA person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.CorroboratedThe same number is printed in 1 other lab document: 1 |
| BrowseCompmax reasoning effort | 83.4% | CheckedA person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Terminal-Bench 2.0max reasoning effort | 67.9% | CheckedA person checked this number against its source on 18 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.CorroboratedThe same number is printed in 1 other lab document: 1 |
| AutomationBenchpublic set · reported later, in DeepSeek-V4-Pro GA Release | 12.8% | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| MMLU-Promax reasoning effort | 87.5% | CheckedA person checked this number against its source on 18 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| LiveCodeBenchmax reasoning effort; problem window not stated | 93.5% | CheckedA person checked this number against its source on 18 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
Sources
Reported by DeepSeek
- DeepSeek V4 Preview Releaseapi-docs.deepseek.com · 24 Apr 2026
Original image - DeepSeek-V4-Pro GA Releaseapi-docs.deepseek.com · 13 Aug 2026
Original image
