Jonathan Kemper / The Decoder:

Alibaba Technical Report: Qwen3-VL beats GPT-5 and Gemini 2.5 Pro on visual tasks and has 100% accuracy on “needle-in-a-haystack” tests for 30-minute videos  —  A few months after launching Qwen3-VL, Alibaba has released a detailed technical report on the open multimodal model.

Jonathan Kemper / The Decoder:

Alibaba Technical Report: Qwen3-VL beats GPT-5 and Gemini 2.5 Pro on visual tasks and has 100% accuracy on “needle-in-a-haystack” tests for 30-minute videos  —  A few months after launching Qwen3-VL, Alibaba has released a detailed technical report on the open multimodal model.

Source: TechMeme
Source Link: http://www.techmeme.com/251130/p3#a251130p3