Qwen: Qwen3.5 397B A17B

Qwen3.5 397B A17B is a text generation model from Qwen, released by the provider on 2026-02-16. It belongs to the Qwen3.5 series and is described as a native vision-language model built on a hybrid architecture that combines a linear attention mechanism with a sparse mixture-of-experts model, which the provider says achieves higher inference efficiency. The model id on api14 is qwen/qwen3.5-397b-a17b, and its main category is text.

The model accepts text, image and video as input and produces text as output. It has a context window of 262144 tokens and a maximum output of 235929 tokens, so long inputs and long generations are both within scope. Because it handles images and video alongside text, it can work directly on multimodal prompts rather than relying on a separate pipeline to convert visual material into text first. Reasoning is also exposed as a parameter, so you can control whether reasoning content is included in the response.

Those capabilities point to a few practical uses. Long document and media review is a natural fit: you can pass a large body of text together with screenshots, diagrams or video frames and ask for a written summary or analysis. Video and image understanding tasks, such as describing footage, answering questions about a clip or extracting structured details from visual material, follow from the input modalities. Text generation work that benefits from a large context, such as multi-file code review, transcript analysis or report drafting, is also supported. Tool calling and structured outputs make it usable in agent-style workflows where the model must return machine-readable results or invoke functions.

You call the model through api14's OpenAI-compatible API using the model id qwen/qwen3.5-397b-a17b. The supported request parameters are frequency_penalty, include_reasoning, logit_bias, logprobs, max_tokens, min_p, presence_penalty, reasoning, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs and top_p. That set covers sampling controls, penalties, deterministic seeding, stop sequences, log probability reporting, reasoning output and tool use, so most standard OpenAI-style request shapes carry over without changes.

Keep the output ceiling in mind: max_tokens can go up to 235929, and the context window is 262144 tokens, so very large prompts leave less room for generation. The model outputs text only, so if you need image or audio output, a different model is required. If your task is purely text and does not need the vision or video inputs, a smaller text-only model may be simpler and faster to work with.