大模型网摘OpenAI Ships GPT-6.1 Terra: Vision and Computer Use for the Mid-Tier, Completing the Astra/Sol/Terra/Luna Lineup
On October 26, OpenAI released GPT-6.1 Terra, and the model landed in Microsoft Foundry a day later. It is the fourth publicly available model in OpenAI's current lineup — and with it, the tiering of GPT-6 Astra (flagship), GPT-6.1 Sol, GPT-6.1 Terra and GPT-6 Luna is now fully on the table.
The headline: visual reasoning and computer use
The release announcement emphasizes three things: fast and efficient, detailed visual reasoning, and precise computer use — plus "clearer, more natural writing." OpenAI positions Terra as a significant step up from GPT-5.6 Terra, aimed at the communication and computer-based tasks that fill the workday: drafting proposals, interpreting technical drawings, filling out forms, editing videos.
The pricing coordinate
GPT-6.1 Terra is priced at $1 per million input tokens and $5 per million output tokens. For reference, that is almost half the price of GPT-5.6 Terra and GPT-6.1 Sol, and around 10x the price of GPT-6 Luna. In other words, Terra sits squarely in the middle: much cheaper than Sol, much more expensive than Luna.
Early takes: strongest at communication and speed
OpenAI's Ted Sanders said in a widely quoted Hacker News comment that Terra is probably better than Astra and Sol at technical communication, similar on vision and computer use but much faster, and a bit worse on most other things. Simon Willison, who ran the same prompts across reasoning levels, noted that Terra occupies the same position as Sonnet in Anthropic's collection — neither the cheapest nor the most expensive, which makes deciding when to use it a genuine trade-off.
For subscribers and API developers
For subscribers, this is not a bombshell release: if you keep hitting usage limits, Terra helps your quota go further, and it is a handy pick for faster computer use or clearer writing. For API developers, it fills the middle slot — smarter than Luna, cheaper than Sol — and automated, repetitive computer-use tasks are exactly the kind of volume work a mid-priced model is suited for.
Where it lands in the cross-comparison
The same price band did not get simpler with Terra's arrival: DeepSeek's V4.1-Flash targets agent cost, while GLM and Kimi each field their own lightweight workhorses. For load-based selection frameworks, see our V4.1-Flash vs GPT-6.1 Sol and GLM-5.3-Flash comparison and the agentic coding cost comparison; for pricing-side reference points, our deepseek-flash and deepseek-v4-pro quick reference covers the official tiers. A companion piece covers Terra's API pricing and access.