GPT-6.1 Terra Cuts Its Price in Half. Developers, Don’t Applaud Too Soon

OpenAI released GPT-6.1 Terra on October 26. Its positioning is straightforward: it can see screens, operate software, and write, while being faster and cheaper.

The official use cases are very specific: drafting proposals, reading technical drawings, filling out forms, and editing videos. Put plainly, these are the jobs that fill the workday and require looking at a screen before taking action.

Pricing is the most concrete part of this release. Standard API pricing is $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens. That is 50% lower than GPT-5.6 Terra and GPT-6.1 Sol. OpenAI’s direction is clear: make developers comfortable handing high-frequency tasks such as reading invoices and entering data into accounting software over to the model.

On benchmarks, Agents’ Last Exam measures a model’s ability to use software in work scenarios. At medium reasoning effort, GPT-6.1 Terra scores 15 percentage points higher than GPT-5.6 Terra at about one-third the cost; it matches GPT-6 Astra’s score at one-tenth the price. At maximum reasoning effort, it is more than 3x more token-efficient than Claude Opus 5.5, costs one-quarter as much per token, and has a task cost about 9x lower. It completes tasks more than 3x faster than GPT-6 Astra and GPT-6.1 Sol.

On the visual reasoning benchmark MMMU-Pro, Terra beats GPT-5.6 Terra in every reasoning setting and comes close to GPT-6 Astra and GPT-6.1 Sol at much lower cost. OSWorld 2.1 tests multi-step desktop application operations; at medium reasoning effort, Terra surpasses GPT-5.6 Terra’s best score at a much lower task cost.

Factuality and coding have also changed. In OpenAI’s internal set of prompts designed to induce errors, at low reasoning effort the error rate drops from 53.6% for GPT-5.6 Terra to 16.7%, a nearly 37 percentage point decrease. At maximum reasoning effort, the error rate falls to 4.3%, close to GPT-6 Astra and GPT-6.1 Sol. On the coding benchmark Terminal-Bench 4.0, maximum reasoning effort solves 50.3% of tasks; GPT-5.6 Terra’s published score is 21.5%. OpenAI also states clearly, however, that the most complex coding tasks should still use GPT-6.1 Sol and GPT-6 Astra.

On safety and usability, Terra reaches the Critical threshold under the cybersecurity category of OpenAI’s Preparedness Framework and retains the safeguards of GPT-6 Astra. OpenAI says it is more likely to stay within task scope, is more willing to explain limitations and blockers, and has strong resistance to prompt injection. The system card is at deploymentsafety.openai.com/gpt-6-1-terra.

It is now available to Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex. When developers call it through the API, the model name is gpt-6.1-terra.

The community reaction is more interesting than the launch copy. One user directly asks whether Terra is better than Sol. The question is not surprising; OpenAI’s naming has already confused people. Others don’t care about faster and cheaper; they just want a model that can beat Opus 5.5. Another complaint hits hard: tokens keep getting cheaper every time, yet the total bill keeps rising.

The most noteworthy thing about this release is that OpenAI has put price-performance front and center. For high-frequency, low-complexity work that requires looking at a screen and clicking buttons, Terra has a clear place. For complex coding tasks, even OpenAI itself does not overclaim. Cheaper tokens usually mean higher call volume, and the bill will not automatically go down. Developers should not rush to applaud; first calculate their own task costs.

Reference link: https://openai.com/index/introducing-gpt-6-1-terra/