DeepSeek v4 Pro
DeepSeekFlagship
ThinkingTool UseStructured Output
About this model
DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. It is designed for advanced reasoning, coding,...
Performance Tier
Flagship
DeepSeek v4 Pro is a flagship model from DeepSeek : the most capable in their lineup.
Best-in-class model from this provider. Highest performance across benchmarks, ideal for demanding tasks.
Pricing
This model is included in Elosia plansAffordable
Low cost. Suitable for sustained use and high-volume interactions.
| Type | per 1M tokens |
|---|---|
| Input (prompt) | $1.17 |
| Output (completion) | $2.34 |
| Cache read | $0.099 |
Capabilities
Context Length1.0M
Max Output Tokens393K
TokenizerDeepSeek
Inputtext
Outputtext
Release DateApril 24, 2026
Benchmarks
General Intelligence
MMLU
90.1%
MMLU-Pro
87.5%
GPQA Diamond
90.1%
Mathematics
MATH-500
Not reported
Programming
HumanEval
76.8%
SWE-bench Verified
80.6%
LiveCodeBench
93.5%
Reasoning
IFEval
Not reported
Humanity's Last Exam
37.7%
Agentic
SWE-bench Pro
55.4%
Terminal-Bench 2.0
67.9%
Where does DeepSeek v4 Pro stand?
Compare its performance index against every other model.
Recommended Use Cases
CodingMathematicsAnalysisResearch
Strengths
- Open-weight MIT — self-hostable frontier model with no vendor lock-in
- Coding leader at its price point (LiveCodeBench 93.5, SWE-bench Verified 80.6)
- MoE 1.6T total / 49B active params — frontier capability with efficient inference
- 1M-token context with hybrid sparse attention (10% of v3.2 KV-cache footprint)
Limitations
- Trails frontier proprietary models on factual recall (SimpleQA-Verified 57.9 vs Gemini 75.6)
- Humanity's Last Exam (37.7) below Claude Opus 4.7 and Gemini 3.1 Pro
- Less extensive safety tuning than Claude/GPT
Frequently asked questions
Resources
This model may use your data for training
