Key points
- Input tokens run $1 per million and output tokens $5, against $2 and $10 for Claude Sonnet 5.5, per The Decoder's comparison.
- Rollout covers paid ChatGPT tiers (Plus, Pro, Enterprise, Edu) via ChatGPT Work and Codex, and developers can call it through the API.
- Artificial Analysis found a 6-point Intelligence Index gain over GPT-5.6 Terra at max effort, with per-task cost of $0.40 versus $1.40.
- On a private benchmark, guessing when the answer is unknown fell to 38 percent from 88 percent for the prior Terra.
Why it matters for marketers: Cheaper per-task agent runs with fewer made-up answers lower the cost of automating drafting, form-filling and reporting, so teams can test workflows that were too expensive before.
OpenAI has released GPT-6.1 Terra, a mid-range model priced at half the API rate of GPT-5.6 Terra. The Information framed the launch as part of a pricing contest with Anthropic. The Decoder reports that Terra undercuts Claude Sonnet 5.5 on both input and output tokens.
OpenAI says the model is aimed at everyday office work, including drafting proposals, reading technical drawings and completing forms, with stronger control of software on a user's behalf. Paid ChatGPT users can reach it now, and developers can use it through the API.
Independent numbers back up the cost story. Artificial Analysis calls it a new high for intelligence per dollar, noting better scores than the previous Terra at less than a third of the cost per task. Hallucination behavior also improved: the model guesses far less often when it lacks an answer, though accuracy rose only slightly. OpenAI also reports gains in prompt-injection defense, relevant for agents that read third-party content.
For marketers and content teams, the practical shift is cost. Routine agent work such as drafting, data entry into web tools and document review gets cheaper to run at volume. Fewer invented answers also reduce the editing burden on generated copy.
Worth watching: OpenAI itself still points to GPT-6.1 Sol and GPT-6 Astra for the hardest coding jobs, so Terra is not a full replacement. Teams should run their own tests on real workflows, and watch whether Anthropic answers with price cuts.
Sources
This brief is our own summary. Read the full reporting at the source.
