Qwen: Qwen2.5 VL 72B Instruct

Qwen: Qwen2.5 VL 72B Instruct is a vision language model from Qwen, the model family developed by Alibaba Cloud. The "72B" in the name refers to the size of the underlying model, and "Instruct" indicates it has been tuned to follow instructions rather than simply continue text. It sits in the text category on api14 and is identified by the provider release date of 1 February 2025. If you have worked with other Qwen models, this one extends the same lineage to inputs that include images alongside text. The provider describes it as proficient at recognizing common objects such as flowers, birds, fish, and insects, and as highly capable of analyzing texts, charts, icons, graphics, and layouts within images.

The model accepts text and image inputs and returns text only. Its context window is 128,000 tokens, and the maximum output is 115,200 tokens, so long documents and lengthy generations both fit comfortably in a single request. The knowledge cutoff is 30 June 2024, which matters when your questions depend on events or releases after that date. Because images are a first class input, you can send a prompt that mixes instructions and visual material, and the model reasons over both when producing its answer. Output is always text, so anything you need in another format has to be produced as text and parsed downstream.

Those capabilities point to a fairly specific set of jobs. Document and chart analysis is the obvious one: feeding screenshots of dashboards, scanned tables, or diagrams and asking for a summary or an extracted value. Visual inspection tasks follow from the object recognition strength, for example classifying or describing what appears in a photo. Layout understanding is useful for processing forms, receipts, and interface mockups where the position of elements carries meaning. And because the context window is large, you can combine several images with a long body of reference text in one call instead of chunking everything yourself.

You call the model through api14's OpenAI-compatible API using the model id "qwen/qwen2.5-vl-72b-instruct". Supported request parameters are frequency_penalty, logit_bias, logprobs, max_tokens, presence_penalty, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, top_k, top_logprobs, and top_p. The presence of response_format and structured_outputs is useful when you need output that conforms to a schema, and seed helps when you want repeatable results across calls. Note that this model is text only in its output, so if your task is speech, audio, or video generation, a different model on api14 will fit better.