OpenAI has released GPT 6.1 Terra, a lightweight artificial intelligence model optimized for desktop automation and computer control. The system handles everyday digital workflows, including reading technical blueprints, filling software forms, and processing invoices. Access is live now for paid ChatGPT tiers, with developer pricing set at $1 per 1,000,000 input tokens.
OpenAI targets system level navigation with this release. The model interacts directly with desktop operating software to execute multi step tasks without human intervention. Common actions include pulling numerical figures from invoices and entering them into accounting software. It also parses visual documents like engineering schematics and assists with editing video files across third party desktop tools.
Benchmark testing highlights gains in both speed and operational efficiency. In the Agents Last Exam evaluation, GPT 6.1 Terra scored 15% higher than GPT 5.6 Terra at medium reasoning while consuming 3x fewer tokens. Under maximum workloads, it matches the performance of the flagship GPT 6 Astra model while running 10x cheaper and 3x faster than both GPT 6 Astra and GPT 6.1 Sol. In the Terminal Bench 4.0 evaluation for command line coding, the model solved 50.3% of tasks at peak reasoning, compared to 21.5% for GPT 5.6 Terra. Internal evaluations also reveal a factual error rate of 16.7% under low reasoning, which drops to 4.3% at maximum reasoning levels.
Safety updates focus on boundary control and adversarial defense. The model resists prompt injections designed to derail automated routines and states its task limits more transparently during execution. OpenAI made the model available to Plus, Pro, Enterprise, and Edu subscribers inside ChatGPT Work and Codex. The API costs $1 per 1,000,000 input tokens and $5 per 1,000,000 output tokens, cutting operational expenses by 50% compared to GPT 5.6 Terra and GPT 6.1 Sol.
