OpenAI has released GPT-6.1 Terra, a new model built to handle complex computer-based work at a lower cost. It focuses on tasks that require visual reasoning and interaction with desktop applications. The release also brings substantial gains in factual accuracy and coding performance over GPT-5.6 Terra.
Terra targets practical workloads that businesses want to automate without paying premium prices for every task. These include processing invoices and interpreting technical drawings. OpenAI says the model can also tackle demanding professional tasks at a fraction of the cost of its more expensive alternatives.
Faster performance at lower costs
GPT-6.1 Terra delivered a significant improvement on Agents’ Last Exam, a benchmark covering complex professional tasks. Challenges include converting architectural drawings into 3D models and documenting individual film shots for restoration teams.
At medium reasoning effort, Terra scored 15 percentage points higher than GPT-5.6 Terra. OpenAI achieved that improvement at roughly one-third of the cost. Terra also matched GPT-6 Astra’s performance on the benchmark at approximately one-tenth of its price per token.
GPT-6.1 Terra: delivering fast and capable computer use at a lower price.
— OpenAI (@OpenAI) October 26, 2026
GPT-6.1 Terra understands what’s on screen, takes action across apps, writes naturally, and is a more capable model for your whole workday with the efficiency to take on more of it. pic.twitter.com/8JTN1yonmV
Speed offers another advantage. OpenAI says Terra completes tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol. At maximum reasoning effort, it also achieves roughly nine times lower task costs than Claude Opus 5.5.
The results extend to desktop automation. OSWorld 2.1 evaluates how well models operate desktop applications through multistep workflows. Terra surpassed GPT-5.6 Terra’s best score at medium reasoning effort while costing substantially less per task.
Terra also performed strongly on MMMU-Pro, which tests visual reasoning. It outperformed GPT-5.6 Terra at every tested reasoning setting. Its scores came close to those of GPT-6 Astra and GPT-6.1 Sol, despite its lower cost.
Fewer errors, better coding
OpenAI has also worked to reduce factual mistakes, which can undermine automated workflows. Its internal benchmark uses prompts designed to expose weaknesses in model responses.
At low reasoning effort, Terra recorded an error rate of 16.7 percent. GPT-5.6 Terra scored 53.6 percent on the same evaluation. That represents a reduction of 36.9 percentage points.
The improvement continued at maximum reasoning effort. Terra’s error rate fell to 4.3 percent, down 18 points from its predecessor. That brings its performance close to the levels reported for GPT-6 Astra and GPT-6.1 Sol.
Coding performance has improved considerably, too. Terminal-Bench 4.0 measures how well coding agents solve command-line problems. One example involves fixing a webpage whose layout shifts as it loads.
At maximum reasoning effort, Terra solved 50.3 percent of the benchmark’s tasks. GPT-5.6 Terra recorded a score of 21.5 percent, giving the new model more than twice its predecessor’s success rate. OpenAI still recommends GPT-6.1 Sol and GPT-6 Astra for its most complex coding challenges.
Availability and API pricing
GPT-6.1 Terra became available on October 27 to Plus, Pro, Enterprise, and Edu users through ChatGPT Work and Codex. Developers can also access it through the OpenAI API using the model identifier gpt-6.1-terra.
Standard API pricing starts at $1 per million input tokens. Cached input tokens cost $0.05 per million, while output tokens cost $5 per million. OpenAI says these rates are 50 percent lower than those for GPT-5.6 Terra and GPT-6.1 Sol.
Terra retains the safeguards developed for GPT-6 Astra. OpenAI reports stronger resistance to prompt injection and better disclosure of limitations. Its preparedness evaluations also place the model at the Critical threshold for cybersecurity.
The lower price could make complex computer automation more accessible to developers and businesses. However, GPT-6.1 Sol and GPT-6 Astra remain the stronger options for the most demanding coding work.