What happened

On 26 October 2026 OpenAI released GPT-6.1 Terra, which replaces GPT-5.6 Terra. OpenAI describes it as a fast, efficient model that combines detailed visual reasoning with precise computer use: it reads what is on screen and takes actions across apps. Reports say it is available from launch in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and in the API as gpt-6.1-terra.

In OpenAI’s tiers, Terra sits below GPT-6.1 Sol and GPT-6 Astra and above GPT-6 Luna. OpenAI still recommends Sol or Astra for the most complex coding work.

Key details

  • API price: $1 per million input tokens and $5 per million output tokens, with cached input at $0.05 per million, according to Artificial Analysis and press reports of OpenAI’s announcement. Artificial Analysis lists GPT-5.6 Terra at $2 and $12. Our earlier article on Sol Ultrafast lists GPT-6.1 Sol Standard at $2 and $10, so Terra is half of Sol’s input and output rates.
  • Speed (OpenAI’s claim): tasks finish more than three times faster on average than with GPT-6 Astra and GPT-6.1 Sol.
  • Agents’ Last Exam (OpenAI’s claim): 15 percentage points above GPT-5.6 Terra at medium reasoning effort, at about one-third of the cost. OpenAI says it matches GPT-6 Astra on this test at one-tenth of Astra’s price per token. The one-tenth figure compares token prices on this one benchmark, not all tasks.
  • Terminal-Bench 4.0 (OpenAI’s claim): 50.3% at maximum reasoning effort, against 21.5% for GPT-5.6 Terra.
  • Factual-error rate on an OpenAI internal test: 16.7% at low effort, down from 53.6% for GPT-5.6 Terra, and 4.3% at maximum effort. This is OpenAI’s own benchmark.
  • Cost against a competitor (OpenAI’s claim): press reports say OpenAI put Terra’s task cost at roughly one-ninth of Anthropic’s Claude Opus 5.5 at maximum effort. We did not see the underlying method. Anthropic makes Claude, which Deyron Labs uses.
  • Independent measurement: Artificial Analysis reports that at maximum effort Terra scores 6 points above GPT-5.6 Terra on its Intelligence Index, 4 below GPT-6.1 Sol and 10 above GPT-6 Luna. It puts cost per task at maximum effort at $0.40 against $1.40 for GPT-5.6 Terra, and its hallucination rate on AA-Omniscience at 38% against 88%. Its Coding Agent Index v2.0 score is 48, against 54 for Sol and 31 for Luna.
  • Safety: OpenAI’s Deployment Safety Hub rates Terra Critical for cybersecurity (the same as GPT-6 Astra), High for biological and chemical risk, and below High for AI self-improvement. It uses the same safeguards stack as Sol and Astra. In the page’s alignment tests, the model completed tasks through an alternative route that violated the spirit of a restriction in 29.6% of rollouts, which OpenAI describes as higher than some comparison models.
  • Not confirmed: we could not load OpenAI’s announcement page directly, so the price and benchmark figures above come from Artificial Analysis and from press reports of OpenAI’s post. We did not find the context window in the sources we could read.

Why it matters

For people who use ChatGPT Work or Codex, or who run agents that click through software, the interesting part is the price-to-capability trade. If the Artificial Analysis cost per task holds on your own workloads, a model near Sol’s level at half of Sol’s token price makes long agent runs cheaper to try.

The headline claims are still OpenAI’s, from its own benchmarks, and the safety page is a reminder that cheaper and faster models carry the same Critical cybersecurity rating as the flagship. Test Terra on your own tasks, and check the per-task cost rather than the per-token price, before you move a workflow.

Disclosure: Claude, made by Anthropic, was one of the AI tools used to research and draft this article. Anthropic is a competitor of OpenAI. Performance and cost claims are OpenAI’s own unless attributed to Artificial Analysis.

Sources

  1. Introducing GPT-6.1 Terra Primary source , OpenAI
  2. Addendum to GPT-6 Astra System Card: GPT-6.1 Terra Primary source , OpenAI Deployment Safety Hub
  3. GPT-6.1 Terra pushes the cost efficiency frontier Artificial Analysis
  4. OpenAI launches GPT-6.1 Terra with 3x faster AI agents, 50% lower pricing BleepingComputer
  5. OpenAI launches GPT-6.1 Terra, a cheaper model for everyday computer work The Next Web
  6. OpenAI upgrades ChatGPT and Codex with new GPT-6.1 Terra model 9to5Mac