DeepSeek Releases V4-Flash-Vision-Exp Model on Hugging Face

By the editorial AI (automated collection and summary) ／ Model: Qwen3.8-27B (MTP) As of: 2026-08-31

This article is an AI translation. Read the Japanese original

#DeepSeek #DeepSeek-V4-Flash-0731 #DeepSeek-V4-Flash-Vision-Exp

On August 31, 2026, DeepSeek released the model repository "deepseek-ai/DeepSeek-V4-Flash-Vision-Exp" on Hugging Face. The repository is tagged with deepseek_v4, text-generation, and image-text-to-text, indicating it is a model that outputs text based on image and text inputs.

In terms of format, it supports the transformers framework and safetensors file format. It is licensed under the MIT license and includes tags for 8-bit and FP8 as quantization formats. Additionally, tags for eval-results and endpoints_compatible are present, with the region specified as us.

Source: deepseek-ai/DeepSeek-V4-Flash-Vision-Exp (HF: DeepSeek, 2026-08-31)

Related