DeepSeek-V4-Pro
deepseek-v4-pro
fields
| nameREQ | DeepSeek-V4-Pro | api-docs.deepseek.com | 2026-05-01 |
| vendorREQ | DeepSeek | api-docs.deepseek.com | 2026-05-01 |
| released_at | 2026-04-24 | api-docs.deepseek.com | 2026-05-01 |
| modality | text, code | api-docs.deepseek.com | 2026-05-01 |
| access | open-weights | api-docs.deepseek.com | 2026-05-01 |
| context_window | 1000000 | api-docs.deepseek.com | 2026-05-01 |
| parameters | 1.6T total / 49B active | api-docs.deepseek.com | 2026-05-01 |
| benchmark_score | 80.6 | deepseekai.guide | 2026-05-02 |
| benchmark_name | SWE-Bench Verified | deepseekai.guide | 2026-05-02 |
| announcement_url | https://api-docs.deepseek.com/news/news260424 | api-docs.deepseek.com | 2026-05-01 |
history · 10 fields · 10 revisions
name1 revision
DeepSeek-V4-Pro
current
DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. 1M context length.
vendor1 revision
DeepSeek
current
DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. 1M context length.
released_at1 revision
2026-04-24
current
DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. 1M context length.
modality1 revision
text, code
current
DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. 1M context length.
access1 revision
open-weights
current
DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. 1M context length.
context_window1 revision
1000000
current
DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. 1M context length.
parameters1 revision
1.6T total / 49B active
current
DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. 1M context length.
benchmark_score1 revision
80.6
current
SWE-Bench Verified: V4-Pro (reasoning_effort=max): 80.6% | LiveCodeBench: 93.5% | Terminal-Bench 2.0: 67.9% (all DeepSeek-reported from the V4 announcement and technical report)
benchmark_name1 revision
SWE-Bench Verified
current
SWE-Bench Verified: V4-Pro (reasoning_effort=max): 80.6% (DeepSeek-reported from V4 technical report)
announcement_url1 revision
https://api-docs.deepseek.com/news/news260424
current
DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. 1M context length.