OpenAI releases GPT-6.1 Terra, a cheaper computer-use model at half the price of GPT-5.6 Terra

OpenAI released GPT-6.1 Terra on October 26, 2026, describing it as a fast, capable computer-use model: it understands what is on screen, takes action across apps and writes naturally. It is rolling out to all Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and developers can use it in the API as gpt-6.1-terra. OpenAI says usage allowances in ChatGPT Work and Codex go further with it than with GPT-5.6.

OpenAI's claims (not independently verified):

• Standard API prices are 50% lower than GPT-5.6 Terra's.
• It substantially outperforms GPT-5.6 Terra on Agents' Last Exam, a test of how well agents use software at work, and matches GPT-6 Astra on it at one-tenth the price per token, finishing tasks over 3x faster on average than both Astra and GPT-6.1 Sol.
• GPT-6.1 Sol and GPT-6 Astra remain OpenAI's most capable models for specialized coding, research and deeper reasoning.
• Standard API pricing is $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens.
• On Terminal-Bench 4.0 at maximum effort it solves 50.3% of tasks, against a published 21.5% for GPT-5.6 Terra. On OpenAI's internal test of error-inducing prompts, the error rate at low effort fell from 53.6% to 16.7%.
• At max effort on Agents' Last Exam, it says Terra uses over three times fewer tokens than Claude Opus 5.5 at a quarter of the price per token, about 9x lower cost per task.
• OpenAI says the model meets the Critical threshold in cybersecurity under its Preparedness Framework and keeps the protections built for GPT-6 Astra; a system card is published.

OpenRouter lists openai/gpt-6.1-terra with a context length of 1,050,000 tokens. OpenAI's model page gives the same 1,050,000-token context window, 128,000 maximum output tokens and a June 1, 2026 knowledge cutoff. It takes text and image input and outputs text; reasoning effort can be none, low, medium (the default), high, xhigh or max; and tool calling requires the Responses API. The page lists cache writes at $1.25 per million tokens (30-minute retention), Batch and Flex at 50% below standard, fast mode at 2x standard, and prompts over 272K input tokens at 2x input and 1.5x output.

Artificial Analysis (a benchmark site, its own measurements) puts the price at $1 per million input tokens and $5 per million output tokens. It scores the model 48 on its Coding Agent Index v2.0, 6 points below GPT-6.1 Sol (54) and 17 above GPT-6 Luna (31), with the xhigh and low effort settings on the cost-efficiency frontier. It reports a 50-point drop in hallucination rate on AA-Omniscience with accuracy up 2 points, and a gain of about 20 points on Terminal-Bench 4.0, which it says now rivals GPT-6.1 Sol at max effort.

Mercor (a benchmark publisher, its own measurements) says GPT-6.1 Terra is OpenAI's top model on its APEX-Agents benchmark, with 68.3% Pass@1 (6th overall) and a 79.3% mean score, 10.1 points more Pass@1 than GPT-5.6 Terra. By domain it reports 80.9% Pass@1 in management consulting (2nd), 66.2% in investment banking (7th) and 57.7% in corporate law.

Arena (a leaderboard publisher, its own measurements) says on October 27 that GPT-6.1 Terra (Max) placed #14 in its Code Arena: WebDev with 1671 points, 151 more than GPT-5.6 Terra, which is now #56. It reports a median cost of $0.40 per task, 28% less than GPT-5.6 Terra. On October 29 Arena added that GPT-6.1 Terra (Max) ranks #8 in Agent Arena across nearly 9,000 agentic sessions from its users, with a +10.6% net improvement and a median cost of $0.62 per task.