Plain-English Summary
A smaller, faster DeepSeek V4 model designed for efficient reasoning and high-volume production.
DeepSeek-V4-Flash
A smaller, faster DeepSeek V4 model designed for efficient reasoning and high-volume production.
DeepSeek V4
Reasoning LLM
GA / open weights
Open weights: Yes
API: Yes
Technical Notes
MoE model with 284B total parameters, 13B active parameters, 1M context, 384K output, and thinking/non-thinking modes.
Best For
Cost-efficient coding, agents, extraction, high-volume reasoning
Watch Out For
Lower capability than V4-Pro; self-hosting still requires capable hardware.
Model Specs
- Maker
- DeepSeek
- Country
- China
- Family
- DeepSeek V4
- Type
- Reasoning LLM
- Release date
- 2026-04-24
- Date confidence
- Exact
- Status
- GA / open weights
- Context window
- 1,000,000 tokens
- Max output
- 384,000
- Parameters
- 284B
- Active parameters
- 13B
- Architecture
- Mixture-of-experts transformer
- Reasoning
- Yes — thinking and non-thinking
- Tool calling
- Yes
- Structured output
- Yes
- API available
- Yes
- Open weights
- Yes
- Self-hostable
- Yes
- License
- DeepSeek model license / repository terms
- Inputs
- Text
- Outputs
- Text
- Approx. price
- See DeepSeek API pricing
- How to access
- Provider website or API
- Model / API ID
- deepseek-v4-flash
- Knowledge cutoff
- Not publicly disclosed
- Fine-tuning
- Self-hosting, fine-tuning, prompting, tools
- Completeness
- 100 %
- Verified
- Jul 18, 2026
- Snapshot date
- Jul 18, 2026
- Source
- Official source