Skip to content

Back to the catalog

OpenAI Launches GPT-6.1 Terra, Its Half-Good AI Model at Half the Price
decrypt.coarticle

OpenAI Launches GPT-6.1 Terra, Its Half-Good AI Model at Half the Price

GPT-6.1 Terra is OpenAI's mid-tier model, released 26 October 2026 with standard API prices 50% below GPT-5.6 Terra. At low reasoning effort its factual error rate on an internal test fell from 53.6% to 16.7%, and at maximum effort it solved 50.3% of Terminal-Bench 4.0 tasks versus 21.5%. It is available in ChatGPT Work and Codex and as gpt-6.1-terra. All figures are OpenAI's own.

TL;DR Reports OpenAI's GPT-6.1 Terra launch, with half-price API rates, lower factual error rates and a Terminal-Bench 4.0 gain over GPT-5.6 Terra.
Document

What it claims

GPT-6.1 Terra is OpenAI's mid-tier model, released 26 October 2026. OpenAI positions it as the everyday model for routine office work done through computer use: reading what is on a screen, clicking through apps and drafting replies. Its standard API prices are 50% below GPT-5.6 Terra. OpenAI says it finishes tasks more than three times faster on average than GPT-6 Astra and GPT-6.1 Sol. For in-depth research or complex coding, OpenAI still directs users to GPT-6.1 Sol or GPT-6 Astra rather than Terra.

Evidence

All figures are OpenAI's own. None are independently reproduced in the article.

The factual-error test is internal and built from prompts designed to trick models into mistakes. At low reasoning effort, the error rate fell from 53.6% for GPT-5.6 Terra to 16.7%. At maximum effort it fell 18 points to 4.3%. Reasoning effort is the setting that controls how long the model thinks, so low-effort and maximum-effort results are not comparable to each other.

On Terminal-Bench 4.0, which scores the share of command-line terminal tasks a coding agent completes, Terra at maximum effort solved 50.3% against 21.5% for GPT-5.6 Terra. No low-effort Terminal-Bench figure is given.

The "three times faster" claim compares Terra against larger models, GPT-6 Astra and GPT-6.1 Sol, not against GPT-5.6 Terra. No accuracy comparison against those larger models is given.

How to use it

Developers call the model through the API as gpt-6.1-terra. In ChatGPT it is rolling out to Plus, Pro, Enterprise and Edu users in ChatGPT Work and in Codex. ChatGPT Work is OpenAI's agent mode for working across apps and websites, and Codex is its coding tool. Free users get the lower-tier GPT-5.6 Luna instead. The 50% cut applies to standard API prices. The article gives no absolute per-token prices.

Limits

The error benchmark is internal and unpublished. Terminal-Bench results are reported only at maximum effort, so the cost of reaching 50.3% is unknown. OpenAI itself does not recommend Terra for complex coding.

Terra first appeared on 26 June 2026 in the GPT-5.6 family, as a limited preview requested by the Donald Trump administration. It opened to the public on 9 July 2026. In September 2026 OpenAI launched GPT-6 Sol and Luna with no new Terra.

Organisations named here

  • OpenAI
  • Anthropic
  • Decrypt
  • Donald Trump
  • Terminal-Bench

Technologies named here

  • GPT-6.1 Terra
  • ChatGPT Work
  • GPT-5.6 Terra
  • Codex
  • GPT-5.6 Luna
  • GPT-6.1 Sol
  • GPT-6 Astra
  • GPT-6 Sol
  • GPT-6 Luna
  • gpt-5.5
  • Claude Opus 5.5
  • ChatGPT
  • OpenAI API

Techniques named here

  • Computer use
  • Reasoning effort