Qwen releases Qwen3.8-2.4T-A95B model weights
Qwen released the model weights and configuration files for Qwen3.8-2.4T-A95B in Hugging Face Transformers format. The 2.4T-parameter MoE model activates 95B parameters per token and is natively designed for 262,144-token context, with extension up to 1,010,000 tokens. Qwen says the artifacts support vLLM, SGLang, and TokenSpeed, while the separate Qwen3.8-Max adds vision input, non-thinking support, a default 1M context length, and built-in tools.
This is a substantial open-weight release, but its claimed capability gains and benchmark results have not been independently verified in the supplied report.
Qwen/Qwen3.8-2.4T-A95B Text Generation • 2.4T • Updated 3 days ago • 26.3k • • 1.38k
library_name: transformers license: other license_name: qwen3.8-max license_link: LICENSE pipeline_tag: text-generation
Qwen3.8-2.4T-A95B
[!Note] This repository contains model weights and configuration files for the post-trained model in the Hugging Face Transformers format. These artifacts are compatible with vLLM, SGLang, TokenSpeed, etc.
[!Tip] For users seeking managed, scalable inference without infrastructure maintenance, the official Qwen API service is provided by Qwen Cloud. In particular, Qwen3.8-Max is the official version based on Qwen3.8-2.4T-A95B with more features, such as vision input & non-thinking support, 1M context length by default, official built-in tools, etc. For more information, please refer to the Qwen3.8-Max Overview.
#Qwen3.8-2.4T-A95B#Qwen3.8-Max#Qwen#Hugging Face Transformers#vLLM#SGLang#TokenSpeed#Qwen Cloud