GPT-6.1 Terra, a new model from OpenAI, is live today in ChatGPT, in Codex and in the API as gpt-6.1-terra, according to OpenAI's announcement. Access in ChatGPT Work and Codex covers Plus, Pro, Enterprise and Edu plans.
Price and access
OpenAI lists the standard API price as $1 per million input tokens and $5 per million output tokens. Cached input costs five cents per million tokens.
- Input: one dollar per million tokens.
- Cached input: five cents per million tokens.
- Output: five dollars per million tokens.
- Where: ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu, plus the API.
OpenAI says that is 50% below GPT-5.6 Terra and GPT-6.1 Sol.
Benchmarks, in OpenAI's words
Every figure here is OpenAI's own, and we found no independent test of them.
- On Agents' Last Exam, OpenAI says GPT-6.1 Terra scores 15 points above GPT-5.6 Terra at about a third of the cost.
- OpenAI says it matches GPT-6 Astra on that test at one tenth the price per token.
- On Terminal-Bench 4.0, OpenAI says it scores 50.3%, against 21.5% in its own comparison.
- At low effort, OpenAI says the hallucination error rate is 16.7%, against 53.6% in its own comparison.
- OpenAI also claims roughly nine times lower cost per task than Claude Opus 5.5 at max effort.
Cyber risk rating
OpenAI says the model meets the "Critical" cyber threshold under its Preparedness Framework.
What this means for you
If you pay for ChatGPT Plus, Pro, Enterprise or Edu, you can try GPT Terra in ChatGPT Work and Codex now. For API users, the lower token price is the checkable part: compare it with your current bill. The benchmark gains are claims to test on your own tasks before switching.
Also note that a cost-per-task comparison depends on how many tokens each model spends, which OpenAI's headline numbers do not show us. For related reading, see OpenAI models arriving in Google Cloud's Model Garden and GPT-6 Pro getting faster multi-agent responses.
