DeepSeek
DeepSeek-V4-Flash
Specs
- Input price
- $0.14/M
- Output price
- $0.28/M
- Context window
- 1,048.576K tokens
A 284B Mixture-of-Experts efficiency model from DeepSeek that activates only 13B parameters per token - top-tier speeds at a fraction of the cost.
Capabilities
- Hyper-lean 13B active parameters, 200+ tokens/second on optimized endpoints
- Persistent reasoning state across tool invocations for long pipelines
- Three toggleable effort modes: Non-think, Think High, Think Max
- Hybrid Attention (CSA + HCA) cuts KV cache to 10% of standard footprint
Best for
- High-volume document pipelines and information extraction
- Conversational search and first-line support with low latency
- Persistent multi-step automation agents
Limitations to keep in mind
- Text-only - does not parse native images or medical scans (DICOM)
- Smaller active parameter scale limits deep factual recall vs flagships
- Can stumble on repository-wide multi-file software engineering
HIPAA-compliant hosting
DeepSeek-V4-Flash is available under our HIPAA-compliant design-partner program, with a signed Business Associate Agreement (BAA), encryption in transit and at rest, and access controls. We are currently onboarding design partners, with general availability coming soon. Run the model behind a unified OpenRouter-style API and swap to another model with a single parameter.
Pricing in context
Open source models like DeepSeek-V4-Flash can be served efficiently on optimized inference infrastructure, with savings passed through to you. Exact savings depend on the model and your volume, but open source inference is typically a fraction of the per-token cost of closed-source frontier models on Azure OpenAI or AWS Bedrock - without cloud egress lock-in or minimum commitments.
Design partner program
Run DeepSeek-V4-Flash under HIPAA
Join the waitlist to be prioritized. We'll reach out with a qualification call and early access.