DeepSeek v4 Flash
DeepSeekBalanced
ThinkingTool UseStructured Output
About this model
DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with 284B total parameters and 13B activated parameters, supporting a 1M-token context window. It is designed for fast inference and...
Performance Tier
Balanced
DeepSeek v4 Flash is a balanced model from DeepSeek : strong performance at a reasonable price.
Strong cost-performance ratio. Reliable for most professional use cases without premium pricing.
Pricing
This model is included in Elosia plansEco
Minimal cost. Ideal for very high volume or simple tasks.
| Type | per 1M tokens |
|---|---|
| Input (prompt) | $0.140 |
| Output (completion) | $0.280 |
| Cache read | $0.0028 |
Capabilities
Context Length1.0M
Max Output Tokens384K
TokenizerDeepSeek
Inputtext
Outputtext
Release DateApril 24, 2026
Benchmarks
General Intelligence
MMLU
88.7%
MMLU-Pro
86.2%
GPQA Diamond
88.1%
Mathematics
MATH-500
Not reported
Programming
HumanEval
69.5%
SWE-bench Verified
79%
LiveCodeBench
91.6%
Agentic
Terminal-Bench 2.0
56.9%
Recommended Use Cases
CodingMathematicsAnalysisGeneral Chat
Strengths
- Outstanding cost-to-performance — ~3x cheaper than v4 Pro for near-frontier reasoning
- MoE 284B total / 13B active — high throughput ideal for agents and coding assistants
- 1M-token context with the same hybrid sparse attention as v4 Pro
- Open-weight MIT — deployable on-prem
Limitations
- Lower factual knowledge density than v4 Pro for recall-heavy tasks
- Max reasoning mode adds significant latency — not ideal for real-time UX
- Smaller MoE — weaker performance on creative writing vs proprietary models
Resources
This model may use your data for training
