DeepSeek-V4-Flash Leaks on Hugging Face: Targets Million-Token Context

TheZachMueller · x · 2026-07-31

A model page for DeepSeek-V4-Flash-0731 has unexpectedly appeared on Hugging Face. Metadata reveals that the new model focuses on Million-Token Context Intelligence and supports FP8 and 8-bit quantization. The repository is currently in an early or private stage with zero downloads.

Related event: DeepSeek-V4-Flash Leaks with Million-Token Context(2 posts)→

Original post →

More from Models

Models channel →