OpenAI has released GPT-6.1 Terra, a mid-range model that costs half as much as the version it replaces. The company is pitching it at the everyday work businesses actually pay for: reading documents, filling in forms, routine coding and operating software on a computer screen.
Terra takes over from GPT-5.6 Terra and sits in the middle of OpenAI’s line-up, above the cheaper GPT-6 Luna and below the more capable GPT-6.1 Sol and GPT-6 Astra. It is available now in the API and to ChatGPT Plus, Pro, Enterprise and Edu users through ChatGPT Work and Codex. OpenAI has not said whether free users will get it.
Cheaper, and better on paper
Through the API, Terra costs $1 per million input tokens and $5 per million output tokens, with cached input at $0.05 per million. That is half the price of GPT-5.6 Terra and of GPT-6.1 Sol. OpenAI says subscribers on ChatGPT Work and Codex plans will be able to complete more tasks within the same usage allowance than with the older or larger models.
The company also claims sizeable quality gains, though the figures come from its own testing. On Terminal-Bench 4.0, a test of command-line coding tasks, Terra solved 50.3 percent of problems at maximum effort, against 21.5 percent for its predecessor. On OSWorld 2.1, which measures how well a model can operate a desktop, it scored 74.2 percent, up from 57.1 percent. On a set of internal prompts designed to trip models into factual mistakes, the error rate at low effort fell from 53.6 percent to 16.7 percent.
OpenAI suggests uses such as drafting proposals, reading technical drawings and entering invoice details into accounting software. It still recommends Sol or Astra for the hardest software engineering work, and says Terra keeps the cybersecurity safeguards built for its top model and resists prompt-injection attacks better than the previous version.
Analysts urge caution. Greyhound Research’s chief analyst, Sanchit Vir Gogia, argued that lower inference prices do not cover the cost of reviewing and integrating what a model produces. In one example he described, a few extra seconds of human checking per task wiped out the savings. Others advise companies to test Terra on their own workloads before switching.
What it means here
For Bangladeshi software firms, outsourcing companies and startups, price is often the deciding factor. API bills are charged in dollars and usually paid by international card, so every cut in per-token cost matters for companies that watch their dollar spending closely. Halving the price of a capable mid-tier model makes it more realistic to automate high-volume back-office work such as invoice processing, document review and customer support triage.
That is also where the risk sits. Much of Bangladesh’s IT and business process outsourcing work involves the same kind of routine tasks that models like Terra are designed to handle. Cheaper automation could help local firms win more contracts with smaller teams, or it could let overseas clients do the work in-house.
The practical advice is the same as the analysts’: run real tests, including on Bangla-language documents and local formats, measure the time staff spend correcting the output, and compare the total cost rather than the price per token alone.
