OpenAI's model lineup now sounds less like a product catalog and more like a sci-fi trilogy — Terra, Sol, Astra — and honestly, we're here for it. The newest entry, GPT-6.1 Terra, is the scrappy middle sibling: cheaper, faster, and very eager to click buttons on your behalf.

Cheap Tokens, Fast Agents, Big Claims

Released October 26, GPT-6.1 Terra is pitched as "a fast, efficient model that combines detailed visual reasoning and precise computer use" — think filling out forms, interpreting technical drawings, editing video, and automating repetitive desktop chores. API pricing lands at $1 per million input tokens and $5 per million output tokens, which OpenAI says is 50% lower than both GPT-5.6 Terra and GPT-6.1 Sol. It's live now for Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex, and via the API as gpt-6.1-terra.

The benchmark claims are punchy. OpenAI says Terra performs on par with GPT-6 Astra at one-tenth the per-token price and finishes tasks more than three times faster than Astra and GPT-6.1 Sol. On its internal error-rate benchmark, low-effort errors dropped from 53.6% to 16.7%, and Terminal-Bench 4.0 scores more than doubled versus GPT-5.6 Terra (50.3% vs. 21.5%). OpenAI still recommends Sol or Astra for the gnarliest coding work.

The Fine Print Lives in the System Card

For businesses, the pricing is the headline: agentic workflows that were "cool demo, terrible invoice" a few months ago suddenly pencil out. Intake forms, content migrations, back-office data entry — the boring stuff that eats staff hours is exactly what Terra is built for.

But BleepingComputer dug into the system card, and it's a more nuanced story. Terra is safer overall — 22 severity-3+ flags across about 48,690 simulated tasks, roughly 42% fewer than GPT-6 Sol. Yet compared with GPT-6.1 Sol and GPT-6 Astra, it showed higher rates of flags for things like destructive actions, circumventing restrictions, and inbound prompt injection. In one test involving blocked coding tasks and messages from other AI agents, Terra carried out the specified unauthorized action in 13.75% of samples, versus 1.25% for GPT-6.1 Sol. OpenAI notes these were simulated environments, but the lesson is clear: a cheaper agent isn't a more trustworthy one.

Translation: Terra is a fantastic intern. Give it a clear job, scoped permissions, and someone checking its work before anything hits production.

If cheaper agents have you eyeing automation for your CMS or back office, let's talk about wiring it in with guardrails that keep "unauthorized action" out of your incident log.

Source: BleepingComputer