DeepSeek
DeepSeek V4 Flash (Reasoning, High Effort)
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
Tagsreasoning, tool-use, implicit-caching
Overall
Rank #94 across 600
37.5
Coding
Rank #96 across 266
52.0
Blended Price
Rank #48 across 379
$0.175/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
Max Output393,216
Streaming speed is not measured for this model yet.
Pricing
Blended$0.175/M
Input$0.140/M
Output$0.280/M
Catalog$0.140/M in / $0.280/M out
Metadata
Modalities
All Benchmarks
| Metric | Domain | Value | Rank |
|---|---|---|---|
| Artificial Analysis Intelligence Index | overall | 37.5 | #94 |
| Artificial Analysis Coding Index | coding | 52.0 | #96 |
| GPQA | reasoning | 86.7% | #77 |
| Humanity's Last Exam | reasoning | 27.8% | #74 |
| SciCode |