OpenAI Shipped Another Model Yesterday: GPT-6.1 Terra at Half Price — First, Check Where It Actually Lives
One-line takeaway: On 26 October 2026, OpenAI released GPT-6.1 Terra: API pricing of $1 per million input tokens and $5 per million output tokens, half the previous generation — but it only appears in ChatGPT Work and Codex, not in regular ChatGPT chats or the free tier.
What you need
- Know which door you are: developers use the API (model ID
gpt-6.1-terra); subscribers use ChatGPT Work or Codex; free-tier users do not get it yet - It is not a free model: "half price" is not "free"
Steps
Step 1: Confirm the launch — announced yesterday, a mid-tier workhorse

PYMNTS reported on 26 October 2026 that OpenAI released GPT-6.1 Terra the same day, positioned for high-volume automation — reading invoices, entering data into accounting software, drafting proposals, interpreting technical drawings, filling out forms, and editing video. OpenAI calls it a significant step up from GPT-5.6 Terra, with standard API prices 50% lower than GPT-5.6 Terra and GPT-6.1 Sol.
Step 2: Do the pricing math — $1/$5, cached input $0.05

Pricing is consistent across sources: $1 per million input tokens, $5 per million output tokens (the previous generation was $2/$10), and $0.05 per million cached input tokens. Third-party evaluator Artificial Analysis reports that at maximum reasoning effort, Terra scores 6 points higher on its Intelligence Index than its predecessor while cost per task falls from $1.40 to $0.40 — a third-party figure, not an OpenAI guarantee.
Step 3: Confirm where to find it — Work and Codex only, not regular chat

Availability matters more than the headline: ChatGPT Work and Codex only, on Plus, Pro, Business, Enterprise, and Edu plans. The free tier is not on the list, and regular ChatGPT conversations do not include it. Some Plus/Pro users may see "Medium" effort as a new default in Power presets. Developers reach it via v1/responses and v1/chat/completions as gpt-6.1-terra. UAE inference residency is not supported.
Costs, caveats, and pitfalls
- Cost: API $1 input / $5 output / $0.05 cached input per million tokens. Subscriber access in Work/Codex depends on plan, region, and workspace settings — a gradual rollout, not same-day for everyone.
- Not a free model: Released yesterday and usable today does not mean free. Free-tier users do not get a Work/Codex door at all.
- Separate benchmark sources: Figures like "15 points higher on Agents Last Exam" or "error rate down from 53.6% to 16.7%" are OpenAI internal benchmarks or third-party evaluations, not your own test results. Do not present them as tested promises in a proposal.
- Do not confuse the names: In the GPT-6.1 family, Terra is the mid-tier model; Sol and Luna shipped earlier in October, Astra is the flagship tier. Pick by price and task, not by the "6.1" version number alone.
Sources
- PYMNTS (26 October 2026): https://www.pymnts.com/news/artificial-intelligence/2026/openai-unveils-half-price-ai-model-to-target-enterprise-automation/
- AI Daily Post (26 October 2026): https://aidailypost.com/news/gpt-61-terra-halves-openai-api
- Availability and pricing explainer: https://www.online-tech-tips.com/openai-gpt-6-1-terra-release-speed-lower-cost/
- Forkast (26 October 2026; screenshot timed out, replaced by AI Daily Post): https://forkast.news/gpt-6-1-terra-arrives-at-1-5-half-the-rate-of-a-model-openai-shipped-27-days-ago/