Qwen has released the open model Qwen3.8 with a 1M token context

The Qwen development team has introduced the large-scale model Qwen3.8-2.4T-A95B-FP8. This new release belongs to the category of open architectures and offers high performance in specialized tasks.

The total size of the model is 2.4 trillion parameters, of which 95 billion parameters are activated during operation. The architecture supports long context processing of up to 1 million tokens, allowing for the handling of massive text arrays and codebases.

For efficient deployment and reduced hardware requirements, it features built-in FP8 quantization. The model demonstrates high results in benchmarks focused on programming and autonomous agents.

The release expands the capabilities for developers to create local and server-based systems with deep context understanding. The source weights and configuration files are available on the repository page.