DeepSeek
DeepSeek V4 Pro (Non-reasoning)
DeepSeek-V4 series incorporate several key upgrades in architecture and optimization: (1) a hybrid attention architecture that combines Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA) to improve long-context efficiency; (2) ManifoldConstrained Hyper-Connections (mHC) that enhance conventional residual connections; (3) and the Muon optimizer for faster convergence and greater training stability
DeepSeek release notesModel Snapshot
TypeLanguage
ReleaseApr 23, 2026
Context1,048,576
Tagsimplicit-caching, reasoning, tool-use
Overall
Rank #144 across 600
31.2
Coding
Rank #63 across 266
59.4
Blended Price
Rank #133 across 379
$0.544/M
Strength Profile
Percentile score by analysis domain.
* Cost is inverted: lower input, output, and blended prices rank higher.
Benchmark Percentiles
Higher bars mean stronger relative placement.
Model Details
Runtime
Output Speed60.1 tok/s
First Token1.12s
Max Output384,000
Pricing
Blended$0.544/M
Input$0.435/M
All Benchmarks
| Metric | Domain | Value | Rank |
|---|---|---|---|
| Artificial Analysis Intelligence Index | overall | 31.2 | #144 |
| Artificial Analysis Coding Index | coding | 59.4 | #63 |
| GPQA | reasoning | 71.7% | #251 |
| Humanity's Last Exam | reasoning | 7.7% | #257 |
| SciCode |