Evergreen Tools
|
← Back to AI Tools

GPT-6.1 Terra: Visual Reasoning and Computer Use

OpenAI's fast model announced 26 October 2026: built for visual reasoning, computer use and writing, aimed at high-volume document and desktop-app work, while the largest GPT-6 models stay the recommendation for the hardest coding

Official Source

This page summarises official announcements — check OpenAI's site for current pricing and availability.

Features

  • ✓ Positioned as a fast model for visual reasoning, computer use and writing, improving on the previous Terra in visual reasoning, factual accuracy and coding
  • ✓ Typical work is high-volume documents and desktop apps: filling in forms, interpreting technical drawings, entering invoice details into accounting software
  • ✓ Accepts text and images and returns text, with a context window of more than 1M tokens
  • ✓ Supports configurable reasoning effort, structured outputs, tool calling, and implicit plus explicit prompt caching
  • ✓ Latency-sensitive work can use fast mode as openai/gpt-6.1-terra-fast; the standard model name is openai/gpt-6.1-terra

How to Use

  1. Start from the task shape: forms, drawings and invoice entry are its home turf, while the hardest coding still belongs to the larger GPT-6 models
  2. Pick openai/gpt-6.1-terra in AI Gateway or an OpenAI-compatible endpoint, and switch to the fast variant when latency matters
  3. Turn on structured outputs to pin the fields down, and feed images in one go for form-style tasks
  4. Enable prompt caching for batch jobs so the same prompt is not billed over and over

FAQ

What is GPT-6.1 Terra?

OpenAI announced GPT-6.1 Terra on 26 October 2026 (official post: https://openai.com/index/introducing-gpt-6-1-terra). It is described as a fast model for visual reasoning, computer use and writing, improving on the previous Terra in visual reasoning, factual accuracy and coding, and pitched at high-volume document and desktop-app work.

How does it divide work with GPT-6.1 Sol?

Per Vercel's AI Gateway listing on 26 October 2026, Terra takes the high-frequency, fast path - filling out forms, interpreting technical drawings, entering invoice details into accounting software - and the same listing says OpenAI still recommends its larger GPT-6 models for the most complex coding tasks. Routine volume to Terra, hard problems to the bigger model.

What are the input and context limits?

The listing states that the model accepts text and images and returns text, with a context window of more than 1M tokens, plus configurable reasoning effort, structured outputs, tool calling and implicit and explicit prompt caching. Note the shape - images in, text out. It is not an image generation or editing model.

How much does it cost?

This page does not publish a price guess. As of writing, billing details for Terra should be taken from OpenAI's own pricing page, because third-party write-ups disagree with each other. Check the official pricing page and your AI Gateway billing dashboard before scaling a batch job onto it.

What is the quickest way to wire it up?

If you already use AI Gateway or an OpenAI-compatible Chat Completions or Responses endpoint, swapping the model name is the whole migration: openai/gpt-6.1-terra for the standard tier, openai/gpt-6.1-terra-fast for latency-sensitive work. It also works with the AI SDK and coding agents connected to the gateway.