TEXT MODEL · Apr 24, 2026

DeepSeek-V4-Flash

By DeepSeek 深度求索

DeepSeek-V4-Flash by DeepSeek: released Apr 24, 2026; 284B total, 13B active parameters; 1M-token context; open weights under MIT.

Verified facts

Outside China
Open weights · verified Sep 29, 2026 · Evidence
Open weights
Yes · verified Sep 29, 2026 · Evidence
License
MIT · verified Sep 29, 2026 · Evidence
Claim status
Company-reported · verified Sep 29, 2026 · Evidence
Parameters
284B total, 13B active · verified Sep 29, 2026 · Evidence
Context window
1,000,000 tokens · verified Sep 29, 2026 · Evidence
List price
Not verified yet
Hugging Face
deepseek-ai/DeepSeek-V4-Flash · verified Sep 29, 2026

Compare current API prices on ModelPriceLab.

Release timeline

  1. DeepSeek releases DeepSeek-V4.1-Flash with 890-byte per-token KV cache
  2. DeepSeek released DeepSeek-V4-Flash-Vision-Exp, its first experimental multimodal model in the V4 family
  3. DeepSeek released DeepSeek-V4-Flash-0731 on Hugging Face under MIT license