GPT-6.1 Terra
Simon Willison’s Weblog Simon Willison ● Covered by 5 sources
OpenAI launched GPT-6.1 Terra, a faster midrange model for visual reasoning and computer use. It’s meant for real workday chores, and its price sits between the cheap and pricey tiers.
Based on reporting by Simon Willison’s Weblog, Simon Willison — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI’s new GPT-6.1 Terra isn’t being sold as the company’s biggest, flashiest model. The pitch is narrower and more useful: better visual reasoning, better computer use, and clearer writing. That framing says a lot. Rather than chasing the usual “best at everything” headline, OpenAI is aiming this one at the dull but constant stuff people actually ask models to do all day.
The examples are familiar office fare: drafting proposals, reading technical drawings, filling out forms, and editing videos. OpenAI says Terra is a significant step up from GPT-5.6 Terra. Ted Sanders, in a Hacker News comment, put it more bluntly: Terra is probably better than Astra and Sol at technical communication, similar on vision and computer use, faster on those tasks, and a bit worse on most other things.
That makes Terra a deliberate middle option. It sits in the same slot as Sonnet in Anthropic’s lineup: not the cheapest, not the most expensive, and therefore harder to pigeonhole. For API developers, that may be exactly the point. It’s the kind of model you reach for when you want something cheaper than Sol but much smarter than Luna, especially for vision, computer use, and communication.
OpenAI also seems to be treating price as part of the product story here. Terra is listed at $1 per million input tokens and $5 per million output tokens, which is almost half the price of GPT-5.6 Terra and GPT-6.1 Sol, and around 10 times the price of GPT-6 Luna. That’s a wide spread, and it explains why Terra feels less like a bombshell and more like a carefully placed rung on the ladder. Subscribers may simply get a little more room before hitting limits. Developers, meanwhile, get a sharper excuse to split workloads by cost and capability.
My take — AI-written commentary, not fact-checked reporting
This is the sensible part of the model boom: not another “greatest ever,” just a tool with a job. The industry keeps pretending every release needs to be a moonshot, but most teams need something that handles forms, drawings, and tedious computer work without setting money on fire. Middle-tier models are where the real product decisions happen, even if the keynote posters prefer the bigger shiny thing.
Read more about this at: Simon Willison’s Weblog
