OpenAI's GPT-6.1 Terra cuts errors by two-thirds at half the price

OpenAI's mid-tier model just got a lot harder to dismiss. GPT-6.1 Terra replaces GPT-5.6 Terra in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users. In the API it is gpt-6.1-terra at $1 / $5 per million input/output tokens, half what GPT-5.6 Terra cost.

The headline number comes from OpenAI's internal benchmark of error-inducing prompts. At low reasoning effort the error rate fell from 53.6% to 16.7%. At maximum effort it reached 4.3%, which OpenAI says nearly matches the far pricier GPT-6 Astra and GPT-6.1 Sol. Coding improved too: Terminal-Bench 4.0 rose from 21.5% to 50.3% at max effort. OpenAI also says Terra finishes tasks more than 3x faster than Astra or Sol. The benchmark is OpenAI's own and hasn't been independently reproduced, so treat the error figures as vendor claims.

The pitch is everyday computer work: sharper visual reasoning and more precise computer use for reading technical drawings, filling out forms and editing video. OpenAI still points the hardest coding jobs to Sol or Astra.

Why it matters: near-flagship reliability at mid-tier prices makes Terra the default for agents doing long, repetitive desktop work.