OpenAI on October 26, 2026 introduced GPT-6.1 Terra, a mid-tier model aimed at visual reasoning, computer use, and everyday office automation. Standard API pricing is $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens—which OpenAI says is 50% lower than the standard rates for GPT-5.6 Terra and GPT-6.1 Sol. The model is available the same day to Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex, and to developers via the API as gpt-6.1-terra.
Key Facts
| Details | |
|---|---|
| Announced | October 26, 2026 (OpenAI blog; OpenRouter lists release date Oct 26, 2026) |
| API model id | gpt-6.1-terra |
| Standard API price | $1 / M input; $0.05 / M cached input; $5 / M output (OpenAI) |
| Price claim vs prior | 50% lower than GPT-5.6 Terra and GPT-6.1 Sol standard API prices (OpenAI) |
| Availability | Plus, Pro, Enterprise, Edu in ChatGPT Work and Codex; OpenAI API |
| OpenRouter listing | Same $1 / $5 rates; Flex $0.50 / $2.50; Fast $2 / $10 (OpenRouter) |
| Terminal-Bench 4.0 (max effort, OpenAI) | 50.3% vs GPT-5.6 Terra 21.5% |
| Internal factual-error rate (OpenAI) | Low effort: 53.6% → 16.7%; max effort: 4.3% |
| Safety (OpenAI) | Meets Critical cybersecurity threshold under Preparedness Framework; system card linked |
| Artificial Analysis Intelligence Index (max) | 48 (vs GPT-6.1 Sol 52, GPT-5.6 Terra 42, GPT-6 Luna 31) |
| AA cost per task (max effort) | $0.40 vs $1.40 for GPT-5.6 Terra |
| AA-Omniscience hallucination (when unsure) | Guesses 38% of the time vs 88% for GPT-5.6 Terra |
Background
Terra sits between OpenAI's cheaper Luna-class and higher Sol / Astra tiers. The company positions GPT-6.1 Terra for high-volume computer-use work—reading invoices into accounting software, drafting proposals, interpreting technical drawings—rather than the hardest coding jobs, for which it still recommends GPT-6.1 Sol and GPT-6 Astra.
On Agents' Last Exam at medium reasoning effort, OpenAI reports a 15 percentage-point gain over GPT-5.6 Terra at about one-third the cost, and performance on par with GPT-6 Astra at roughly one-tenth its price per token. At max effort it claims Terra is more than three times as token-efficient as Claude Opus 5.5 at one-fourth the price per token (~9× lower cost per task) and finishes tasks over three times faster than GPT-6 Astra and GPT-6.1 Sol. On MMMU-Pro and OSWorld 2.1 it says Terra beats GPT-5.6 Terra across reasoning settings at lower cost. Those comparative figures are OpenAI's; Artificial Analysis independently places max-effort Terra on its Intelligence Index and Coding Agent Index cost-efficiency charts.
Reactions and Impact
Artificial Analysis called GPT-6.1 Terra a push of the cost-efficiency frontier: at max effort it gains 6 Intelligence Index points over GPT-5.6 Terra for less than a third of the cost per task, with larger gains (7–13 points) at lower effort, and a Coding Agent Index v2.0 score of 48. THE DECODER summarized the same OpenAI pricing and benchmark claims and noted Terra as a mid-range rival to Anthropic's Sonnet tier at half Sonnet 5.5's cited token price. OpenRouter listed the model the same day at the official $1 / $5 rates.
What We Don't Know Yet
- Whether third-party agent benchmarks (beyond Artificial Analysis and OpenAI's published tables) will match OpenAI's Agents' Last Exam, OSWorld, and Claude Opus 5.5 cost-per-task comparisons.
- Full system-card details beyond the cybersecurity Critical threshold and the prompt-injection rates THE DECODER attributed to the card (99.99% direct / 98.8% indirect)—readers should check the live system card.
- How ChatGPT Work and Codex rate limits translate into "more tasks per plan" in practice; OpenAI states the qualitative claim without publishing a conversion table.
- Cache-write pricing is cited by Artificial Analysis at $1.25 / M; OpenAI's launch post lists input, cached input, and output only.