GitHub radar
Qwen3.8-Flash-Next runs locally via GGUF
Unsloth released a GGUF build of Alibaba's multimodal Qwen3.8-Flash-Next — the model that understands both text and images is now runnable on a laptop via Ollama.
Unsloth packaged Alibaba's Qwen3.8-Flash-Next as quantized GGUF files for local deployment. The original model, built by the Qwen team, is positioned as an architectural preview of what Qwen4 will bring — unlike ordinary multimodal models that simply describe images, it reasons over visuals as part of an extended conversation. Native context window is 262K tokens, extensible to 1M. Handles agentic reasoning, software engineering tasks, office automation, and image-based question answering. Runs via Ollama, llama.cpp, or LM Studio — pick the quantization level that fits your hardware.
Why a vibe-coder should care
The original Qwen3.8-Flash-Next was server-only in FP8 format. This GGUF build makes it runnable on a regular laptop — no cloud, no API fees. If you work with screenshots, design files, or illustrated documents and want a local model that can reason about them in context, this is now a practical option.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Install Qwen3.8-Flash-Next locally via Ollama using Unsloth's GGUF: https://huggingface.co/unsloth/Qwen3.8-Flash-Next-GGUF — pick the right quantization for my hardware and show me how to send it an image along with a question.
A regular laptop with 16 GB of RAM or more — the 4-bit version takes around 8–10 GB (exact size depends on the variant you choose).
Open on Hugging Face▌ More finds