Qwen3.7 Flash is Alibaba's cheapest multimodal model, a native vision-language reasoning model released July 27, 2026. It accepts text, images, and video within a ~991K-token context window and returns up to 64K tokens of text. At $0.03 input / $0.13 output per million tokens, it is among the least expensive multimodal models available anywhere, positioned for high-volume agent loops, classification pipelines, and document processing where cost efficiency outweighs peak capability.