Unsloth releases DeepSeek-V4-Flash-Vision-Exp model

unsloth · hf · 2026-08-31

Unsloth has released a new model unsloth/DeepSeek-V4-Flash-Vision-Exp.

Based on deepseek-ai/DeepSeek-V4-Flash-Vision-Exp, this release includes a quantized version. The model uses the text-generation pipeline, is compatible with Transformers and Safetensors, is tagged as conversational, licensed under MIT, and supports endpoints.

Original post →

More from Multimodal

Multimodal channel →