Skip to content
What is GPT-6.1 Terra? OpenAI's $1 model for images and screens
Back to Blog
#AI models#GPT-6.1 Terra#gpt-6.1-terra#GPT-6.1 Terra pricing#GPT-6.1 Terra API#OpenAI

What is GPT-6.1 Terra? OpenAI's $1 model for images and screens

Zohaib Masood

GPT-6.1 Terra is OpenAI's new mid-priced model for work that starts with an image or a screen and ends with an action: reading a scanned form, matching a product photo to a catalogue entry, clicking through an app, then writing up the result. It launched on 26 October 2026 in the OpenAI API as gpt-6.1-terra and in ChatGPT Work and Codex. Through the API it costs $1 per million input tokens and $5 per million output tokens.

If you run high-volume document, screen or catalogue jobs through a more expensive model, test Terra on them first. For open-ended planning and the hardest reasoning, keep a larger model in charge.

What does it need to run in your product?

An OpenAI API account on a paid usage tier, and the Responses API if you want tools. These are the requirements in OpenAI's model page, read on 28 October 2026:

  • Model id: gpt-6.1-terra. It takes text and images in and returns text. It does not take audio or video, and it cannot be fine-tuned.
  • Tools: tool calling requires the Responses API. There, it supports function calling, structured outputs, web and file search, code interpreter, computer use and MCP.
  • Reasoning: reasoning.effort accepts none, low, medium (the default), high, xhigh and max.
  • Size: a 1,050,000-token context window, up to 128,000 output tokens, and a knowledge cutoff of 1 June 2026.
  • Rate limits: the free tier is not supported. The lowest paid tier, Build, allows 5,000 requests and 1 million tokens a minute.

Two limits matter for regulated work. OpenAI's release notes say Terra does not support UAE inference residency, based on where inference runs rather than where you are. If you need data kept in a region, Microsoft offers it through Foundry in Data Zone Standard deployments in the US, EU and APAC, at a small premium: $1.10 in and $5.50 out per million tokens in the US zone.

What is it good at, and what is it not for?

OpenAI describes it as:

A fast, efficient model for visual reasoning, precise computer use, and clear, natural writing.
— OpenAI model documentation

The examples OpenAI and Microsoft give are routine business tasks: interpreting technical drawings, filling in forms, drafting emails and proposals, checking campaign images, matching supplier photos to catalogue listings, and acting on application screens. Microsoft's launch post sets out the division of labour with OpenAI's larger model: GPT-6 Astra plans open-ended, multi-step work, and Terra handles the high-volume visual and screen-based tasks inside it at a lower price.

The launch benchmarks are OpenAI's own. Before you switch a production job, run your own test set through Terra and the model you use now, and compare completion rate, error rate and cost per finished job, not cost per token.

What does a job cost?

Take an invoice-reading job that sends 4,000 input tokens (the scanned page plus instructions) and gets back 1,000 output tokens, reasoning included. At OpenAI's standard prices, read on 28 October 2026:

  • Input: 4,000 × $1 ÷ 1,000,000 = $0.004
  • Output: 1,000 × $5 ÷ 1,000,000 = $0.005
  • Per invoice: $0.009. For 10,000 invoices a month: $90.

The same job on Claude Sonnet 5.5, at the $2 and $10 on Anthropic's pricing page read the same day, costs $0.018 an invoice, or $180 a month. If the work can wait, OpenAI's Batch and Flex prices are 50% lower, so the Terra job drops to $45.

Watch three things in the price list. Prompts over 272,000 input tokens are billed at twice the input rate and 1.5 times the output rate for the whole request. Cached input costs $0.05 per million, so repeated instructions get cheap. Regional processing adds 10%.

Can your team use it in ChatGPT?

Yes, on paid plans, but not in ordinary chat. OpenAI's release notes say Terra is rolling out in ChatGPT Work and Codex to Plus, Pro, Business, Enterprise and Edu plans, and is "available in Work and Codex, not Chat". Plan usage limits are unchanged. Some Plus and Pro users will see Terra at Medium become the default in Work's Power presets, in place of GPT-6.1 Sol at Light. API use is billed separately from any ChatGPT subscription.

Want GPT-6.1 Terra built into your product or workflow, and tested against the model you use today? See what we build or tell us about the job.

Sources