Back to Flash News

OpenAI launches GPT-6.1 Terra, touting computer operation abilities and lower API prices

OpenAI posted on X that it has launched GPT-6.1 Terra, which focuses on computer operation abilities and can understand screen content and perform actions across applications. Compared with GPT-5.6 Terra, the new model’s standard API price is 50% lower.

The company said that in the Agents’ Last Exam test, which measures an agent’s ability to use office software, its performance is on par with GPT-6 Astra, but its per-token price is only one-tenth that of the latter, and its average task completion speed is more than 3 times that of GPT-6 Astra and GPT-6.1 Sol. GPT-6.1 Terra is being rolled out starting today to Plus, Pro, Enterprise and Edu users of ChatGPT Work and Codex, and developers can also use it through the API.

The OpenAI system card addendum states that GPT-6.1 Terra is a fast and efficient model designed for high-volume professional work, with strong image analysis and computer use capabilities, and it uses the same safety protections as GPT-6.1 Sol and GPT-6 Astra.

Original sources:

  1. Title: GPT-6.1 Terra: delivering fast and capable computer use at a lower price. (official post) Publisher: OpenAI (@OpenAI) Publication date: 2026-10-26 URL: https://x.com/OpenAI/status/2114780711664398349 Source nature: original post by the party involved Verification status: read in this instance (2026-10-26) Supporting content: launched GPT-6.1 Terra; can understand screen content and perform actions across applications; a more efficient model for a full workday

  2. Title: Addendum to GPT-6 Astra System Card: GPT-6.1 Terra Publisher: OpenAI Publication date: 2026-10-26 URL: https://deploymentsafety.openai.com/gpt-6-1-terra Source nature: official system card Verification status: read in this instance (2026-10-26) Supporting content: GPT-6.1 Terra is a fast and efficient model designed for high-volume professional work, with strong image analysis and computer use capabilities; it uses the same safety protections as Sol and Astra

  3. Title: OpenAI has released GPT-6.1 Terra, pushing the cost efficiency frontier. (analysis post) Publisher: Artificial Analysis (@ArtificialAnlys) Publication date: 2026-10-26 URL: https://x.com/ArtificialAnlys/status/2114781877202768360 Source nature: third-party analysis Verification status: read in this instance (2026-10-26) Supporting content: per-token price is half or less of GPT-5.6 Terra ($1/$5 vs $2/$12); cached input price reduced to one-quarter

ABAB AI Insight

OpenAI’s launch of GPT-6.1 Terra is positioned as an efficiency model for high-volume professional work, with a particular emphasis on computer use abilities, including understanding screen content and performing actions across applications. This capability directly addresses enterprise users’ needs for agents to complete office software tasks, and differentiates it from Astra, which in the GPT-6 series emphasizes frontier capabilities.

The pricing disclosed by the company shows that the standard API price is 50% lower than GPT-5.6 Terra. In the Agents’ Last Exam test, its performance is on par with GPT-6 Astra, but its per-token price is only one-tenth of the latter. Its average task completion speed is said to be more than 3 times that of Astra and GPT-6.1 Sol. These comparisons all come from the company’s own test data and have not yet been fully verified by independent third parties.

The model is being opened to Plus, Pro, Enterprise and Edu users of ChatGPT Work and Codex starting today, while API access is also available. The system card addendum explicitly states that it uses the same safety protection stack as GPT-6.1 Sol and GPT-6 Astra, and it has been evaluated as a high-capability model in areas such as cybersecurity. The combination of pricing and speed makes the model more suitable for batch professional work scenarios that frequently require computer operation capabilities.

Compared with the previously released GPT-6.1 Sol, Terra further lowers cost efficiency while maintaining comparable computer use performance. Developers can call it directly through the API, and enterprise users can use it in existing Work and Codex environments without waiting for a new product form.

ABAB News · Cognitive Law

  1. After the price of computer operation capabilities drops to one-tenth of Astra’s level, the access threshold for high-volume professional work scenarios is significantly lowered.

  2. Within the same family of models, the efficiency-oriented version and the frontier-capability version form a clear division of labor in task speed and per-token cost.