GPT-6.1 Terra in its Max configuration ranked eighth on Agent Arena, a leaderboard focused on tasks in which a model acts on behalf of a user. It recorded a 10.6% net improvement and a median task cost of $0.62, while broader evaluations placed its cost range between $0.40 and $0.62. The model scored 50.3% on Terminal-Bench 4.0, compared with 21.5% for GPT-5.6 Terra, and also achieved 60% on Agents’ Last Exam and about 74% on OSWorld. Priced at $1 per million input tokens and $5 per million output tokens, it is roughly 50% cheaper than GPT-6.1 Sol and GPT-5.6 Terra and reportedly completes tasks more than three times faster than its costlier counterparts.
Source: Read the original article


