GPT-6.1 Terra: OpenAI's new mid-tier model halves the API price and doubles Terra's coding score
Quotable summary
GPT-6.1 Terra is OpenAI’s new fast, low-cost model for computer use, visual reasoning and everyday writing. It went live on 26 October 2026 in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and in the API as gpt-6.1-terra. Standard API pricing is $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens. OpenAI says that’s 50% lower than GPT-5.6 Terra and GPT-6.1 Sol. On OpenAI’s own numbers, Terra scores 15 points higher than GPT-5.6 Terra on the Agents’ Last Exam computer-use benchmark at medium effort, and it more than doubles the old Terra’s Terminal-Bench 4.0 coding score (50.3% vs 21.5%). OpenAI also rates it “Critical” for cybersecurity, so it ships with the safeguards built for GPT-6 Astra. Translation: the cheap model is now good enough to click through your spreadsheets for you, and it costs half as much.
What changed
| Item | Detail |
|---|---|
| Model | GPT-6.1 Terra (API ID gpt-6.1-terra), successor to GPT-5.6 Terra |
| Released | 26 October 2026 |
| Where | ChatGPT Work and Codex (Plus, Pro, Enterprise, Edu); OpenAI API |
| API price | $1.00 / 1M input, $0.05 / 1M cached input, $5.00 / 1M output |
| vs. siblings | Standard API prices 50% lower than GPT-5.6 Terra and GPT-6.1 Sol (OpenAI) |
| Context | 1,050,000-token context window; max 922,000 input and 128,000 output tokens; knowledge cutoff 1 Jun 2026 (API docs) |
| Modalities | Text and image in; text out. No audio or video |
| Reasoning effort | none, low, medium (default), high, xhigh, max; tool calling requires the Responses API |
| Safety tier | Meets the Critical cybersecurity threshold under OpenAI’s Preparedness Framework; keeps GPT-6 Astra’s protections |
OpenAI’s benchmark claims
All figures below are OpenAI-reported and haven’t been independently reproduced yet.
| Benchmark | GPT-6.1 Terra claim |
|---|---|
| Agents’ Last Exam (computer use), medium effort | +15 points over GPT-5.6 Terra at about one-third the cost; “on par” with GPT-6 Astra at one-tenth its price per token |
| Agents’ Last Exam, max effort vs Claude Opus 5.5 | >3x as token-efficient, one-fourth the price per token, ~9x lower cost per task |
| Speed | Finishes tasks >3x faster than GPT-6 Astra and GPT-6.1 Sol |
| Terminal-Bench 4.0, max effort | 50.3%, vs GPT-5.6 Terra’s published 21.5% |
| Error-inducing prompts (internal), low effort | Error rate 16.7%, down from 53.6% |
| Same benchmark, max effort | Error rate 4.3% (down 18 points) |
| Terminal-Bench-Science 0.1, max effort | Beats GPT-6.1 Sol; within five points of Opus 5.5 at <1/5 the estimated cost |
Why it matters
- Agent economics just shifted. Computer-use agents that read invoices, fill out forms or work through desktop apps burn lots of tokens. If a model priced like a mid-tier option really matches a frontier model on those tasks, a lot of automation that didn’t pencil out last week might now. OpenAI’s own example is reading incoming invoices and entering them into accounting software.
- Know where it stops. OpenAI itself says GPT-6.1 Sol and GPT-6 Astra remain the better choice for the hardest coding and in-depth research. Terra is the workhorse, not the flagship.
- “Critical” on a cheap model is the real headline. Under OpenAI’s Preparedness Framework, Critical is the top cybersecurity risk tier. Putting a Critical-rated model at a $1 input price means cheap access to strong offensive-security capability. That’s why OpenAI says Terra keeps the Astra-level protections. Expect defenders and red-teamers to test those guardrails quickly.
- Benchmarks are vendor numbers. Agents’ Last Exam and the error-inducing-prompt set are OpenAI-run evaluations. Treat the comparisons, especially the one against Anthropic’s Claude Opus 5.5, as claims until third-party leaderboards catch up.
FAQ
How much does GPT-6.1 Terra cost in the API? $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens at standard rates, per OpenAI’s launch post.
Can I use it in ChatGPT? Yes, in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users. OpenAI says the same usage allowance covers more tasks with Terra than with GPT-5.6 Terra, and substantially more than with GPT-6.1 Sol or GPT-6 Astra.
Should I switch my coding agent to Terra? For routine command-line and app-automation work, it’s worth a trial: OpenAI reports 50.3% on Terminal-Bench 4.0, more than double the old Terra. For the hardest software engineering, OpenAI still recommends Sol or Astra.
What does the “Critical” cybersecurity rating mean? It’s the highest risk tier for cyber capabilities in OpenAI’s Preparedness Framework. OpenAI says Terra keeps the protections developed for GPT-6 Astra and has strong resistance to prompt injection in its tests. Details are in the system card.
Sources
| Source | Role | Link |
|---|---|---|
| OpenAI | Primary: launch post, pricing, benchmarks, safety tier | OpenAI — Introducing GPT-6.1 Terra |
| OpenAI API docs | Primary: model ID, context window, token limits, modalities | OpenAI API — GPT-6.1 Terra model page |
| OpenAI Developer Community | Primary: launch announcement and availability | OpenAI Community — GPT-6.1 Terra announcement |