GPT-6.1 Terra

OpenAI has released GPT-6.1 Terra, a model the company describes as fast and efficient, built around detailed visual reasoning, precise computer use and clearer, more natural writing. The announcement, as Simon Willison notes, leaned heavily on those two capabilities rather than on raw reasoning scores. Terra is positioned as a significant step up from GPT-5.6 Terra.
The naming puts Terra in OpenAI's current frontier lineup alongside GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna. Willison places Terra in the same slot that Sonnet occupies in Anthropic's model collection: neither the cheapest nor the most expensive tier. That middle position is convenient for pricing but awkward for buyers, since the decision between adjacent tiers is rarely obvious from a spec sheet.
Pricing is the clearest signal of intent. Terra costs $1 per million input tokens and $5 per million output tokens. That is almost half the price of GPT-5.6 Terra and of GPT-6.1 Sol, and roughly ten times what GPT-6 Luna costs. For teams running high-volume workloads, the gap between Terra and Sol is large enough to change architecture decisions; the gap between Terra and Luna is large enough that Luna remains the obvious choice for cheap, simple tasks.
OpenAI's Ted Sanders offered a candid tiering guide in a comment on Hacker News. He said Terra is probably better than both Astra and Sol at technical communication, similar on vision and computer use but much faster, and a bit worse on most other things. For API developers he framed it as a midrange option: cheaper than Sol but considerably smarter than Luna, particularly for vision, computer use and communication tasks. For consumer subscribers he was more measured, describing it as less of a headline release — useful if you are hitting usage limits, or if you want clearer writing or faster computer use.
That framing is worth taking seriously. Computer use — an agent driving a browser or desktop interface through screenshots and synthetic clicks — is token-hungry in a way chat rarely is. An agent that screenshots after every action burns input tokens continuously, and a single task can involve dozens of steps. A model at a fifth of Sol's price that is roughly comparable on computer use and faster to boot is a plausible default for repetitive automation, even if it trails on harder reasoning. The trade-off is that cheaper per-step execution only pays off if the agent's success rate stays high; a model that misfires more often can cost more in retries and cleanup than it saves per token.
Willison also published a pelican test at each reasoning level, a long-running informal benchmark in which a model is asked to draw a pelican riding a bicycle. The "max" reasoning setting produced a pelican that cost 11.3161 cents and took 4 minutes and 11 seconds. He paired this with a comparison grid covering every reasoning level of Astra, Sol, Terra and Luna. Such tests say little about production reliability, but they do expose latency and cost behavior at the top of the reasoning range — over four minutes for a single image is a meaningful constraint for anything interactive.
Why it matters: developers choosing a default model for agentic and vision-heavy work now have a cheaper, faster option between Luna and Sol, and the price gap is wide enough that many teams will reasonably try Terra first and escalate only when it fails. The main open question, which only real workloads will answer, is whether its computer-use accuracy holds up well enough that the savings survive contact with retries. For product teams, the practical change is a lower cost floor for browser automation and document-heavy workflows; for subscribers, it is a modest quality-of-life option rather than a reason to switch.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
GPT-6.1 Terra It's always interesting to note which aspects of a new model are emphasized in the release announcement. In the case of GPT-6.1 Terra, OpenAI went with visual reasoning and computer use: We’re introducing GPT-6.1 Terra , a fast, efficient model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing. A significant step up from GPT‑5.6 Terra, it excels at the communication and computer-based tasks that fill the workday, from drafting proposals and interp