Qwen3.8-2.4T-A95B Hugging Face Model CardProvider report
2026-08-12Based on revision 207bd685a7e3696cfaff12ded7c6a7ea0f88c996 of the Qwen/Qwen3.8-2.4T-A95B Hugging Face model card, captured on August 12, 2026. This snapshot includes all 32 text evaluations and 151 available scores; it does not copy the 62 multimodal metrics from the Qwen3.8-Max release report. The source score column remains labeled Qwen3.8-Max. The same model card states that Qwen3.8-Max is based on this open model and adds vision input, non-thinking mode, a default 1M context, built-in tools, and other capabilities. The text scores published with the open model are therefore associated with Qwen3.8-2.4T-A95B while the original column label is retained in raw.source_model_label, without attributing the managed model's extra capabilities to the open model. '--' cells do not create scores.