Qwen3-VL Plus

QwenCloud's marketplace page documents the qwen3-vl-plus-2025-12-19 snapshot, naming the exact version based on a December 19, 2025 cut. It announces the model will be retired on October 10, 2026, a near-term lifecycle event developers should plan around. The page reports pricing of $0.20 per 1M input tokens and $1.60 per 1M output tokens at the ≤32k tier, with a 262K context window, 258K max input, and 32K max output. It claims superior reasoning, analysis, and style control versus the September 23 snapshot, plus lower latency and faster responses, with support for text, image, and video inputs.

Vals AI lists Qwen 3 VL Plus as an open-weight Alibaba multimodal model released September 23, 2025, with a 262K-token context window, 33K max output tokens, and token pricing of $0.20 input / $1.60 output per million. Input modalities include text and image, with video and file input unsupported. The page reports an average benchmark accuracy of 50.29% and latency of 51.81 seconds across the Vals evaluation set. The benchmark breakdown includes MedCode at 31.65% (rank 91/105), MedScribe at 77.13% (rank 69/107), and SAGE at 42.10% (rank 54/90), placing Qwen 3 VL Plus in the middle tier among open-weight models on Vals medical benchmarks. Default hyperparameters are temperature 0.7 and max output 32,768, with the default provider listed as Alibaba. This gives developers a third-party performance reference point for the Qwen3-VL-Plus variant.

Alibaba Cloud Model Studio's official release notes document two Qwen3-VL-Plus entries: an October 2025 "Qwen3-VL-Plus Invocation Parameter Update" and a January 2026 "Qwen3-VL-Plus and Qwen-Plus Model Update." Both are listed as first-party model changelog items for the Qwen3-VL-Plus variant specifically. The excerpt confirms the model receives recurring parameter and capability updates through Alibaba's managed Model Studio platform. The release-notes page is a first-party Alibaba documentation index, placing these updates on the official provider changelog rather than third-party catalogs. Exact usage in the changelog reflects ongoing maintenance of the Qwen3-VL-Plus endpoint within Model Studio, with an invocation parameter change in late 2025 preceding a broader model update in January 2026. These are the only model-specific announcements for Qwen3-VL-Plus surfaced from official Alibaba sources in this candidate set.