Alibaba

Qwen3.8-Max is Alibaba Qwen's hosted 2.4-trillion-parameter mixture-of-experts flagship for coding, professional work, multimodal analysis, and long-horizon agents. Effective September 5, 2026 (UTC+8), the qwen3.8-max endpoint automatically serves the September 2 snapshot qwen3.8-max-0902 (alias qwen3.8-max-2026-09-02). QwenCloud says 0902 improves coding depth, multi-tool agent delivery, and visual understanding while keeping the 1M context, thinking mode, tools, and unchanged billing.

API ID qwen3.8-max

Context tokens

1,000,000

Qwen3.8-2.4T-A95B is Qwen's downloadable 2.4-trillion-parameter sparse mixture-of-experts model with 95B activated parameters. The official model card documents text-only input and output, mandatory thinking, 262,144 native context extensible to roughly 1.01M, configurable reasoning effort, and a dedicated Qwen3.8-Max license.

Native context

262,144

Qwen3.8-27B is Qwen's deployment-oriented dense multimodal model for coding, professional work, research, and long-horizon agents. The official repository documents 27B parameters, native image and video understanding, flexible reasoning effort, a 262,144-token native context extensible to 1M, and compatibility with Transformers, vLLM, SGLang, and TokenSpeed.

Native context

262,144

Qwen3.8-Flash-Next is Qwen's experimental open-weight preview of the architecture intended to underpin Qwen4. The official Hugging Face card documents a 125B MoE with 6B activated parameters plus 51B n-gram embeddings and a 4B MTP head, native 262,144-token context extensible to 1M, and text, image, and video input.

Native context

262,144

Qwen-Image-3.0 is Alibaba's third-generation image model for image generation, editing, complex layouts, and multilingual text rendering. The official Qwen release highlights inputs up to 4.5K, small text rendering down to 10 pixels, and native support for twelve languages.