OpenAI has officially launched GPT-6.1 Terra, a specialized reasoning model engineered for visual processing, computer use, and automated document tasks. Available in the API as gpt-6.1-terra, the model replaces GPT-5.6 Terra and features a 1,050,000-token context window with up to 128,000 output tokens. Standard pricing is set at $1.00 per million input tokens and $5.00 per million output tokens—half the API cost of GPT-6.1 Sol and one-tenth that of GPT-6 Astra.
Designed for automated agentic tasks, Terra operates via OpenAI’s Responses API and supports tools including hosted shell, code interpreter, and computer use. On the Agents’ Last Exam v1 benchmark for professional computer tasks, Terra achieved a 57.6% score at medium reasoning effort ($0.29 per task), matching GPT-6 Astra’s accuracy score at a fraction of Astra’s $4.82 per-task cost. OpenAI reports the model runs over three times faster than both Astra and Sol.
While Terra shows significant gains on visual evaluation metrics such as MMMU-Pro (85.5%), OpenAI noted that higher-tier models like Sol and Astra remain recommended for the most complex coding benchmarks like Terminal-Bench 4.0. The new model is currently available to API developers and across ChatGPT Plus, Pro, Enterprise, and Edu tiers.
Why it matters
OpenAI is aggressively driving down per-token reasoning and computer-use costs, making high-volume agent automation commercially viable.
Developers can now execute complex visual and interface automation tasks at 3x faster speeds compared to previous generation frontier models.
Tiered API pricing models allow operators to optimize costs by reserving top-tier frontier models strictly for complex code generation.
Source: marktechpost.com



