Alibaba / Qwen

Alibaba's biggest Qwen3.5 model with over 1 trillion parameters. Supports thinking mode with 82K chain-of-thought output. Part of the Qwen3.5 series featuring hybrid linear attention + MoE architecture. Optimized for reasoning, agentic tasks, RAG, and tool calling.

Qwen3.5 native vision-language Flash model with hybrid linear attention and sparse MoE architecture. 1M context with ultra-low pricing, a major leap forward over the Qwen 3 series in inference efficiency.

Qwen3.5 series 397B-A17B native vision-language model built on a hybrid architecture integrating linear attention with a sparse mixture-of-experts model. State-of-the-art performance with higher inference efficiency.

Qwen3.5 Plus offers strong all-around performance across academia, finance, marketing, programming, and science domains. 1M context with competitive mid-tier pricing.

Alibaba's flagship 235B MoE model with hybrid thinking/non-thinking modes. Top open-source model on most benchmarks.

Alibaba's top open-weights model. Hybrid thinking/non-thinking mode. Multilingual.

Open reasoning model matching o1-mini on math and coding. Apache 2.0 licensed.

Alibaba 72B vision-language model. Understands images, hour-long video, documents, and structured data.

Best open-source code model at 32B size. Matches GPT-4o on coding benchmarks.

Alibaba's frontier open model. Strong coding, math, and multilingual capability.