Qwen3.8-Flash-Next is an experimental AI model from Alibaba's Qwen team for answering questions and generating text, with support for thinking and tool calling in compatible applications. This app runs Unsloth's UD-IQ4_XS quantized edition locally using Strata.

Strata is an open-source inference engine for running large AI models locally. It coordinates GPU memory, system RAM, and storage so the full model does not need to fit in GPU memory. Model Console connects the engine to Olares.

**Key features**

- Generate text responses with support for thinking and tool calling in compatible applications.
- Use Model Console to download the model, view its status, and adjust engine settings.
- Connect compatible applications through Olares Router or the OpenAI-compatible chat API.

**This edition**

This edition accepts text input only; image and video input are disabled. The default context limit is 131,072 tokens, with one request processed at a time. Available context depends on memory and engine settings.

**Before first use**

First startup downloads approximately 94 GB of model files plus additional files needed by Strata, then prepares the model. Model Console may show the main download as complete while Strata is still preparing the model.

An AVX2-capable CPU and an NVIDIA driver from the R580 series or newer are required. The preset is intended for a host with 96 GiB of RAM. See the listed resource requirements before installing.

[Model documentation](https://huggingface.co/Qwen/Qwen3.8-Flash-Next) · [Quantized model](https://huggingface.co/unsloth/Qwen3.8-Flash-Next-GGUF) · [Strata documentation](https://github.com/Niko1221/Strata)