Qwen3.5 Plus

The SemiAnalysis InferenceX evaluation page for Qwen3.5-397B-A17B — the open-weights checkpoint underlying EmpirioLabs' Qwen3.5 Plus — documents it as a native vision-language foundation model with 397B total and 17B activated parameters per forward pass, released under Apache 2.0. It describes Qwen3.5-Plus as the host Independent evaluation placed Qwen3.5-397B-A17B at an Artificial Analysis Intelligence Index score of 45, ranking third among open-weights models at publication, with a GDPval-AA agentic ELO of 1,221 versus 860 for Qwen3 235B. The checkpoint unifies thinking and non-thinking behavior in a single artifact, operating in

A Baidu Baike encyclopedia entry on the Qwen3.5 series confirms the release on February 16, 2026, listing both "Qwen3.5-Plus" and "Qwen3.5-397B-A17B" as the flagship variants — directly corroborating the subject model name. It adds deployment details: the open-source distribution also includes a 2B dense model and a 35 The entry reports an API price of 0.8 RMB per million tokens (framed as 1/18th of Gemini 3 Pro's price) and an MMLU-Pro knowledge-reasoning score of 87.8 for Qwen3.5-Plus, with visual capabilities covering multimodal reasoning, text recognition, and video understanding. It also notes Day 0 adaptation by NVIDIA, AMD, Ap

Alibaba officially open-sourced Qwen3.5 on February 16, 2026, with the first release being Qwen3.5-397B-A17B, which the announcement explicitly also names "Qwen3.5-Plus" — the same model EmpirioLabs serves as its Qwen3.5 Plus endpoint. The model is described as a natively multimodal foundation model covering text, imag The release emphasizes inference efficiency through a combination of linear attention and a sparse mixture-of-experts design, with 397B total parameters and only 17B activated per forward pass. Alibaba positions Qwen3.5-397B-A17B as targeting reasoning, coding, agent capabilities, and multimodal understanding while low