DeepSeek-V4-Flash-0731 (Byalifiction mirror)

A Hugging Face mirror/re-upload of DeepSeek's efficiency-tier V4 Flash model (0731 release) — a 284B-parameter Mixture-of-Experts LLM with ~13B active parameters per token and a 1M-token context window, optimized for coding, chat, and agentic workflows. The canonical model is published by DeepSeek under the deepseek-ai organization; this Byalifiction copy is a third-party re-hosted artifact. Source: supabase.