DeepSeek V4 Flash by DeepSeek
DeepSeek V4 Flash is an efficiency-focused Mixture-of-Experts model—284B parameters total with 13B active—plus a 1M-token window and very low per-token prices. It is a strong open-weight option when you want speed and cost control without giving up long context.
Specs
| Context window | 1.05M tokens |
|---|---|
| Max output | 943.7K tokens |
| Input modalities | text |
| Output modalities | text |
| Price (input) | $0.091 / 1M tokens |
| Price (output) | $0.182 / 1M tokens |
| Web search | Yes |
| Vision | No |
| Thinking | Yes |
Related models
Last updated 2026-10-31 from the CompareIntel model registry and an OpenRouter spec snapshot dated 2026-10-25. Pricing and context figures follow OpenRouter.