OpenAI: GPT Luna Latest

GPT Luna Latest is a text model from OpenAI. It is published under the model id ~openai/gpt-luna-latest on api14, and OpenAI describes it as always redirecting to the latest model in the GPT Luna family. That means the id is a moving target: when OpenAI updates the Luna family, requests to this id follow the newest member rather than staying pinned to a specific snapshot.

On the input side the model accepts text, images and files, and it produces text only. The context window is 1,050,000 tokens, which is large enough to hold entire codebases, long document sets or extended conversation histories in a single request. The maximum output is 128,000 tokens, so long generations are possible without chaining calls. Because images and files are accepted alongside text, you can send screenshots, scanned pages or attachments together with your instructions and get a single text response.

Those capabilities point to a few practical patterns. Large context plus file input suits document analysis, such as summarising contracts, extracting fields from reports or answering questions across a set of PDFs. Image input makes it usable for reading charts, diagrams, screenshots and photographed pages. The reasoning parameters and tool support make it a reasonable choice for agent style work, where the model calls functions, inspects results and continues. Structured outputs and response format handling fit pipelines that need JSON or a fixed schema rather than free prose.

You call it through api14's OpenAI-compatible API using the model id "~openai/gpt-luna-latest". The supported request parameters are include_reasoning, max_completion_tokens, max_tokens, reasoning, reasoning_effort, response_format, seed, structured_outputs, tool_choice, tools and verbosity. In practice that covers token limits for the reply, control over how much reasoning effort is spent, deterministic sampling via seed, schema constrained output, and tool or function calling with a choice of tool selection behaviour.

One thing to keep in mind is the redirect behaviour. Since the id resolves to whatever is currently the latest GPT Luna model, output can shift when OpenAI ships an update, which matters if you depend on reproducible results. If you need a fixed snapshot, a pinned model id is the safer option. The model is text only on output, so anything requiring generated images or audio needs a different model.