GPT-6.1 Terra Is $1 In and $5 Out and Meets OpenAI's Critical Cyber Threshold. Artificial Analysis Scores It 48; the Astra-Parity and 9x-Cheaper-Than-Opus Claims Are OpenAI's.
OpenAI shipped GPT-6.1 Terra on October 26 for ChatGPT Work and Codex, the API, Amazon Bedrock, and Azure Foundry, at $1 per million input tokens and $5 output. The product post rates it Critical on cybersecurity under the Preparedness Framework, with ExploitBench at 97.10% and SEC-Bench Pro at 83.33%. OpenAI claims parity with GPT-6 Astra on Agents' Last Exam at a tenth of the price and about 9x lower cost per task than Claude Opus 5.5; both are vendor numbers. Artificial Analysis scores it about 48 on the Intelligence Index, 48 on Coding Agent Index v2.0 against Sol's 54, $0.40 per Index task against $1.40 for GPT-5.6 Terra, and hallucination down from 88% to 38%.
By T.W. Ghost
OpenAI shipped GPT-6.1 Terra today, October 26, 2026, and the price and the safety rating arrive in the same post. The price: $1 per million input tokens, $5 per million output, $0.05 per million cached input, on the API as gpt-6.1-terra, in ChatGPT Work and Codex for Plus and above, and same day on Amazon Bedrock and Azure Foundry. The rating: Critical on cybersecurity under OpenAI's Preparedness Framework, with biological and chemical capability at High. The deployment-safety page prints the numbers behind the rating: ExploitBench ACE-normalized mean of 97.10% and SEC-Bench Pro pass@1 of 83.33%.
The claims OpenAI makes about capability are vendor claims, and we say so each time: on par with GPT-6 Astra on Agents' Last Exam at roughly a tenth of Astra's per-token price; about 9x lower cost per task than Claude Opus 5.5 at max effort; about 3x the speed of Astra and Sol. OpenAI also says, in the same post, that GPT-6.1 Sol and Astra remain the better choice for the most complex coding. Artificial Analysis has already scored it, which is where the independent numbers come from. We read the product post, the deployment-safety page, AA's article and model page, and the AWS announcement. Here is what held up.
What You Are Calling
| GPT-6.1 Terra | |
|---|---|
| Released | October 26, 2026 |
| Where | ChatGPT Work and Codex (Plus, Pro, Enterprise, Edu); API gpt-6.1-terra; Amazon Bedrock and Azure Foundry same day |
| Price | $1 in / $5 out / $0.05 cached input, per 1M tokens (OpenAI) |
| Context | 1,000,000 tokens on AA's model card; not stated in the product post we fetched |
| Preparedness rating | Cybersecurity Critical; biological and chemical High (OpenAI) |
| Cyber evals | ExploitBench 97.10%; SEC-Bench Pro 83.33% (OpenAI system card) |
| OpenAI's own guidance | Sol and Astra remain better for the hardest coding |
Critical is in the product post, not a footnote. OpenAI put the rating next to the price, and the deployment page carries the two cyber benchmarks that justify it. That is the context for every cheap-and-fast line below.
Confirm the context window on OpenAI's docs. One million is on Artificial Analysis's card. The product post body we read did not state it. Check before you quote it as OpenAI's number.
The Benchmarks, Sorted by Who Reported Them
| Claim | Number | Who |
|---|---|---|
| Agents' Last Exam: on par with GPT-6 Astra at about one-tenth the price per token | parity | Vendor (OpenAI) |
| Cost per task vs Claude Opus 5.5 at max effort | about 9x lower | Vendor (OpenAI; token efficiency times price) |
| Speed vs Astra and Sol | about 3x | Vendor (OpenAI; not rechecked) |
| ExploitBench ACE-normalized mean | 97.10% | OpenAI (system card, deployment-safety page) |
| SEC-Bench Pro pass@1 | 83.33% | OpenAI (same) |
| Intelligence Index | about 48 | Independent (Artificial Analysis) |
| Coding Agent Index v2.0 | 48, six below GPT-6.1 Sol at 54 | Independent (AA) |
| Cost to run the Intelligence Index | about $0.40, vs about $1.40 for GPT-5.6 Terra | Independent (AA) |
| AA-Omniscience hallucination rate | 38%, vs 88% for GPT-5.6 Terra | Independent (AA) |
Four things to keep straight:
- The Astra comparison is OpenAI's, and the Terra comparison is AA's. OpenAI claims Astra parity on one exam. AA's $0.40 against $1.40 is against the previous Terra, GPT-5.6 Terra, not Astra. Do not let the two get spliced into "a tenth the price of Astra per AA."
- The hallucination drop is also versus GPT-5.6 Terra. 88% to 38% is the same-tier predecessor, not GPT-5.4 and not Sol.
- 48 is the independent number, and Sol is 54. On AA's Coding Agent Index v2.0 Terra lands six points under Sol, which matches OpenAI's own advice to use Sol or Astra for the hardest coding.
- The 9x Opus figure is a vendor cost-per-task claim at max effort, built from OpenAI's token-efficiency measurement times list price. We have covered how those comparisons move with effort settings; treat this one the same way until an independent cost-per-task run exists.
Verified vs Unconfirmed: The Scorecard
| Claim | Verdict |
|---|---|
Released Oct 26; ChatGPT Work and Codex; API gpt-6.1-terra; Bedrock and Foundry same day | Verified (OpenAI; AWS what's-new) |
| $1 / $5 / $0.05 cached per 1M tokens | Verified (OpenAI product post) |
| Cybersecurity Critical; bio and chem High | Verified (OpenAI product post and deployment-safety page) |
| ExploitBench 97.10%; SEC-Bench Pro 83.33% | Verified as OpenAI's measurements (system card) |
| Astra parity on Agents' Last Exam; about 9x cheaper per task than Opus 5.5; about 3x faster | Vendor |
| AA Intelligence Index about 48; Coding Agent v2.0 48 vs Sol 54; $0.40 vs $1.40; hallucination 38% vs 88% | Verified (Artificial Analysis) |
| 1M context as an OpenAI-stated fact | Unverified (on AA's card; confirm on OpenAI docs) |
| Bedrock list price equals $1 / $5 | Unverified (confirm on AWS pricing) |
Who Should Use It
Use Terra for high-volume agent loops where you were paying GPT-5.6 Terra prices: AA's independent read is the same tier of intelligence at $0.40 per Index task against $1.40, with hallucination at 38% against 88%.
Skip it for the hardest coding, where OpenAI itself points you to Sol or Astra, and skip quoting the Astra-parity or 9x-Opus lines without the word "vendor" in front of them. If your deployment review cares about Preparedness ratings, the Critical cyber label is on the box.
The frontier is elsewhere: Sol is still 54 to Terra's 48 on the coding board, and the cyber capability that earned the rating is the same reason OpenAI moved its own training offline this month. Our model comparison has the current standings side by side.
Update, October 29, 2026
- Oct 27: the index changed scale. Artificial Analysis replaced v4.3.2 with Intelligence Index v5 the day after this launch. The 48 above was read from AA's model page the morning of the changeover, and the $0.40 against $1.40 comparison comes from AA's launch-day article on the old scale. On v5, AA lists GPT-6.1 Terra (max) at $0.96 per Index task, the cheapest of the top ten it shows, with GPT-6.1 Sol (max) at 55 and GPT-6 Astra (max) at 57. Re-read the live row for Terra's v5 score before quoting the 48.
Sources
- OpenAI: introducing GPT-6.1 Terra
- OpenAI deployment safety: GPT-6.1 Terra
- Artificial Analysis: GPT-6.1 Terra pushes the cost-efficiency frontier
- Artificial Analysis: GPT-6.1 Terra model page
- AWS what's new: OpenAI GPT-6.1 Terra on Amazon Bedrock
- Artificial Analysis: Intelligence Index v5
*Not sure whether the cheap tier or the coding tier fits your workload? Take the free 2-minute quiz and get matched. Then read about the model OpenAI says to use when Terra is not enough: GPT-6.1 Sol.*