OpenAI: GPT-6 Astra Pro (batch)

GPT-6 Astra Pro is a text generation model from OpenAI, listed in the api14 catalog as OpenAI: GPT-6 Astra Pro (batch) under the model id openai/gpt-6-astra-pro:batch. It is the same underlying model as GPT-6 Astra, served with reasoning.mode set to pro, a configuration aimed at higher-quality responses on complex tasks. OpenAI published the model on 4 September 2026.

The model accepts text, images and files as input and returns text. Its context window is 1,050,000 tokens, and it can produce up to 128,000 tokens in a single response. The large input window and the generous output budget are the two numbers that shape most decisions about it: you can send long documents, attached files and images together in one request, and still have room for a substantial generated answer rather than a short summary. Output is text only, so anything you need back must be expressed as text, including structured data.

Those capabilities point to a fairly specific set of jobs. Long-context review work fits well, such as reading across a large document set, a technical specification or a bundle of source files and answering questions that depend on details spread throughout. Mixed input helps when a task combines screenshots, scanned pages or diagrams with written material. The support for tools and tool_choice makes it usable as the reasoning step in an agent loop, and response_format plus structured_outputs let you constrain replies to a schema your code can parse directly. For work that needs a consistent sampling seed, seed is available.

To call it, use api14's OpenAI-compatible API and pass openai/gpt-6-astra-pro:batch as the model. The supported request parameters are max_tokens, reasoning, reasoning_effort, include_reasoning, verbosity, response_format, structured_outputs, tools, tool_choice and seed. max_tokens controls the length of the generated response, reasoning and reasoning_effort govern how much internal reasoning the model performs, include_reasoning determines whether that reasoning is surfaced to you, and verbosity influences how expansive the written answer is. Tools and tool_choice drive function calling, while response_format and structured_outputs handle schema-constrained replies.

Two limits are worth weighing. The output modality is text only, so if your pipeline needs audio, images or other generated media, this model is not the right fit. And because it runs in pro mode for higher-quality responses on complex tasks, simpler requests may be better served by GPT-6 Astra, which shares the same underlying model without that setting.