“The Qwen 3 VL vision-language model has achieved performance on language tasks that is on par with the Qwen language-only models.”