OpenAI releases GPT-6.1 Terra: a cheap model for computer work

Terra is a fast, economical model that combines detailed visual reasoning with precise computer control and more natural writing. It clearly surpasses GPT-5.6 Terra — it handles everyday computer work from drafting documents through reading technical drawings to filling forms and editing video.

On the Agents' Last Exam benchmark, which tests professional tasks (such as converting an architectural drawing into a 3D model), Terra scores 15 percentage points higher than its predecessor at medium effort — for roughly a third of the price. It matches Astra's performance at a tenth of the cost per token. It completes tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol.

OpenAI invested heavily this year in reducing hallucinations. On the internal error-prompt benchmark, the error rate at low effort dropped from 53.6% (GPT-5.6 Terra) to 16.7%, and at maximum effort to 4.3% — nearly on par with Astra and Sol.

The jump in programming is similarly pronounced: on Terminal-Bench 4.0 Terra solves 50.3% of command-line tasks, more than double its predecessor's score (21.5%). For the hardest code, Sol and Astra remain the better choice.

Standard API pricing is $1 per million input tokens and $5 per million output tokens — 50% less than GPT-5.6 Terra and GPT-6.1 Sol. The model has been available since launch for all paid tiers in ChatGPT Work and Codex, and for developers via the API as gpt-6.1-terra.