DeepSeek-V4-Flash
Open Weights
DeepSeek
The efficient tier of DeepSeek's V4 generation (April 2026): a 284B total / 13B active MoE with a 1M-token context window and the same hybrid attention design as V4-Pro. MIT-licensed for low-cost, high-throughput deployments.
Strengths
- Strong capability-per-cost with only 13B active parameters
- 1M-token context with efficient hybrid attention
- MIT license - free commercial self-hosting
- Far more deployable than V4-Pro on modest hardware
Caveats
- Lower ceiling than V4-Pro on the hardest tasks
- Preview release - behavior may change before stable
- Vendor
- DeepSeek
- Context Window
- 1,000,000 tokens
- License
- MIT
- Release Date
- 2026-04-24
- Modalities
- textimage
Capabilities
Vision
Audio
Video
Tool Use
Resources
Pricing
Open weights under the MIT License - free to self-host; low managed pricing via DeepSeek and other providers.