Monday, 22 June 2026 | Updating Daily AI insight, written for builders

DeepSeek V4-Flash

DeepSeek V4-Flash — Specifications

DesenvolvedorDeepSeek
TipoLLM (MoE)
ModalidadeText → Text
Parâmetros284B total / ~13B active (MoE)
Janela de contexto1M
Saída máxima384K
LicençaMIT (open)
Pesos abertos✅ Yes
Lançamento2026-04
Input price$0.14 /1M
Output price$0.28 /1M
API providersDeepSeek, OpenRouter

🖥️ Run it locally

VRAM (4-bit)~140 GB
Minimum GPU2× H100 80GB (4-bit)

Official page →

The lighter member of the DeepSeek V4 family — 284B total / ≈13B active, a 1M-token context, open MIT weights, at ultra-low pricing for high-volume use.

Scroll to Top