OpenAI launches GPT-6.1 Terra for scaled computer use, claiming ~50% lower API prices
OpenAI released GPT-6.1 Terra, a fast model aimed at visual reasoning, precise computer use, and clearer writing for workday tasks. It claims standard API prices 50% below GPT-5.6 Terra and GPT-6.1 Sol ($1 / $0.05 cached / $5 per 1M tokens) and availability in ChatGPT Work and Codex for Plus/Pro/Enterprise/Edu plus the API as gpt-6.1-terra. Benchmarks and cost multiples are OpenAI claims; Preparedness treats it as Critical in cybersecurity.
OpenAI introduced GPT-6.1 Terra as a fast, efficient model combining detailed visual reasoning and precise computer use with clearer, more natural writing—a step up from GPT-5.6 Terra for workday tasks such as drafting proposals, reading technical drawings, filling forms, and editing video. OpenAI says Terra advances the cost–intelligence frontier, with standard API prices 50% lower than GPT-5.6 Terra and GPT-6.1 Sol for high-volume automation; in ChatGPT Work and Codex the same allowance covers more tasks than GPT-5.6 Terra and substantially more than GPT-6.1 Sol or GPT-6 Astra. It is available starting today to Plus, Pro, Enterprise, and Edu users in Work and Codex, and via the API as `gpt-6.1-terra`, at $1 per million input tokens, $0.05 cached input, and $5 output.
On capability, OpenAI reports several benchmark claims: on Agents' Last Exam at medium reasoning effort, +15 points vs GPT-5.6 Terra at about one-third the cost, and parity with GPT-6 Astra at roughly one-tenth its price per token; at max effort it claims >3× token efficiency vs Claude Opus 5.5 at one-fourth the price (~9× lower cost per task) and >3× faster than Astra and Sol. On MMMU-Pro it says it beats GPT-5.6 Terra at every effort and nearly matches Astra/Sol; on OSWorld 2.1 at medium effort it says it tops prior Terra at a fraction of the cost. On an internal error-inducing set, low-effort error rate fell from 53.6% to 16.7% and max-effort to 4.3%. On Terminal-Bench 4.0 at max effort it claims 50.3% solved vs 21.5% published for GPT-5.6 Terra, while recommending Sol/Astra for the hardest coding. Treat scores and multiples as OpenAI claims.
For safety, the blog says GPT-6.1 Terra meets the Critical cybersecurity threshold under the Preparedness Framework and keeps protections developed for GPT-6 Astra; alignment evals claim better scope-keeping and honest disclosure of blockers, with strong prompt-injection resistance. The Oct 26 System Card addendum on the Deployment Safety Hub states Terra is treated as Critical in cybersecurity and High for Biological and Chemical capability, using the same safeguards stack as GPT-6.1 Sol and GPT-6 Astra (with model-compatible probes). Detailed tables are in that addendum; independent verification remains ongoing.
Sources
- Introducing GPT-6.1 Terra — OpenAI
- Addendum to GPT-6 Astra System Card: GPT-6.1 Terra — OpenAI Deployment Safety Hub
- GPT-6.1 Terra — OpenAI Developers