GPT-6.1 Terra Released: Agent Benchmark Improves by 15 Points Over 5.6 Terra, at About One-Third the Cost
GPT-6.1 Terra has been released, focused on building faster and cheaper agents that can reason over images and operate a computer. It scores 15 points higher than 5.6 Terra on Agents' Last Exam, at about one-third the cost of the latter. API pricing is $1/M for input, $0.05/M for cached input, and $5/M for output, and it is gradually rolling out in the paid plans of ChatGPT and Codex.
Why read this
SelectedAI score 84
The original text provides the new model's improvement on the agent benchmark and its API pricing, which can be used to assess its cost and capability positioning.
Topics
Body
GPT-6.1 Terra is here!! 🚀
build faster, more affordable agents that can reason over images and use computers.
+15 points vs 5.6 Terra on Agents’ Last Exam at roughly 1/3 the cost.
API pricing: $1/M input, $0.05/M cached, $5/M output.
Rolling out across paid plans in ChatGPT & Codex ✨