OpenAI ships GPT-6.1 Terra to ChatGPT and Codex: 3x faster agent runs at half the token price

Terra was left out of last month's GPT-6 Sol and Luna release, per 9to5Mac, and this update closes that gap. GPT-6.1 Terra replaces GPT-5.6 Terra across ChatGPT Work, Codex and the API, available to Plus, Pro, Business, Enterprise and Edu tiers.

The pitch is agent economics. Terra is tuned for native computer use — driving desktops and browsers, reading screens, taking actions — and OpenAI says it completes multi-step enterprise workflows more than three times faster than GPT-6 Astra. On OpenAI's Agents' Last Exam benchmark it improves 15 points over GPT-5.6 Terra. Token pricing drops 50% relative to GPT-5.6 Terra, explicitly to encourage high-frequency autonomous use. Context runs up to 1 million input tokens.

Independent signals are mixed but useful. LlamaIndex's Jerry Liu placed Terra 5th on ParseBench for document parsing: strong on tables, ahead of GPT-6 Sol but behind Claude Opus 5.5 and GPT-6 Astra, and weaker on layout, charts and formatting. At around 1 cent per page, he called it a reasonable option for table-heavy documents.

Distribution is broad: Terra reached GA on Amazon Bedrock and Microsoft Foundry the same week. The open question is reliability. Speed and cost compound for agents, but so do errors. Watch for real-world computer-use failure rates and how Terra stacks up against Anthropic's computer-use models.