NEWS · MODELS · #2602
Introducing GPT-6.1 Terra
A new AI model named GPT-6.1 Terra was announced, positioned as bringing fast and capable computer use to everyone. The announcement emphasizes improved speed and general-purpose utility, but the source provides limited technical detail in the headline.
KEY POINTS
- A new AI model named GPT-6.1 Terra was announced, positioned as bringing fast and capable computer use to everyone.
- The announcement emphasizes improved speed and general-purpose utility, but the source provides limited technical detail in the headline.
- A release of a new GPT-6.1-branded model matters because it indicates further advancement and potential wider deployment of large generative models that can change how people use computers.
DECISION BRIEF
CONFIDENCE
WHAT CHANGED
OpenAI announced GPT-6.1 Terra, a new mid-tier model positioned as faster and more cost‑efficient for everyday computer tasks and automated software control.
WHY NOW
The announcement includes immediate availability and reported large reductions in API token prices and per‑task costs, which could change cost/performance tradeoffs for users and developers now.
WHO IS AFFECTED
Directly affected: ChatGPT Plus, Pro, Enterprise, and Edu users (in ChatGPT Work and Codex) and developers using OpenAI's API; organizations using mid‑tier models for office automation, visual reasoning, or desktop agents are likely to consider switching or testing the model.
CONFIRMED
• OpenAI announced GPT-6.1 Terra with the tagline "Bringing fast and capable computer use to everyone." (OpenAI excerpt). • The Decoder reports OpenAI released GPT-6.1 Terra and halved API prices versus GPT-5.6 Terra. • Reported token prices in the Decoder excerpt: input tokens $1 per million, output tokens $5 per million; cached inputs $0.05 per million. • Decoder excerpt states GPT-6.1 Terra is positioned for office tasks and automated software control and is available to Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex and via the API. • Decoder excerpt cites benchmark claims: intelligence index +6 points over GPT-5.6 Terra at maximum reasoning effort; cost per task examples ($0.40 vs $1.40). • Decoder excerpt gives comparative benchmark scores and metrics (Coding Agent Index v2.0 score 48; AA‑Omniscience guessing rate 38% vs 88% for GPT-5.6 Terra; relative performance statements on MMMU‑Pro and other benchmarks) as reported in that article.
UNCERTAIN
Missing or unclear in the supplied excerpts: • OpenAI's primary excerpt is brief and provides no technical specifications, model size, training data, or latency figures. • The Decoder's claims rely in places on private or third‑party benchmarks and proprietary analyses (e.g., "Artificial Analysis," AA‑Omniscience, private benchmarks) but the excerpts do not show methodology, sample sizes, or independent reproductions. • The Decoder excerpt is truncated (OSWorld 2.1 result is incomplete), so full benchmark context is missing. • It is not clear which claims are OpenAI's official measurements versus independent reporter analysis; attribution within the excerpt is sometimes to "OpenAI" and sometimes to external analysts. • No regional availability, rate‑limit, or long‑term pricing commitment details are present in the excerpts.
WHAT TO WATCH
Concrete observable signals to watch next: 1) Full OpenAI announcement text and technical appendix (model specs, latency, training/data notes). 2) OpenAI public pricing page and API changelog for exact token pricing, tier limits, and caching rules. 3) Official OpenAI blog or docs confirming availability windows and region/plan rollout. 4) Independent benchmark reports reproducing AA‑Omniscience, MMMU‑Pro, Coding Agent Index, and OSWorld results with methodology. 5) Third‑party tests on hallucination/accuracy rates and end‑to‑end task cost measurements. 6) Developer forum / GitHub / Stack Overflow activity showing adoption, integration issues, or performance reports. 7) Competitor responses (pricing or model announcements from Anthropic, others) and any enterprise contract announcements referencing Terra. 8) API usage telemetry/layer‑1 metrics if published (latency, error rates, throughput) and pricing/usage terms changes.
WHY IT MATTERS
A release of a new GPT-6.1-branded model matters because it indicates further advancement and potential wider deployment of large generative models that can change how people use computers.
EVIDENCE MAP
4Editorial claims linked to specific sources, with support, contradiction and context shown separately.
OpenAI announced a model named GPT-6.1 Terra and described it as 'Bringing fast and capable computer use to everyone'.
SUPPORTEDChecked 2026-10-26 · 2 supporting
The Decoder reports GPT-6.1 Terra is a mid-range model and that OpenAI halved its API prices compared with GPT-5.6 Terra.
SUPPORTEDChecked 2026-10-26 · 1 supporting
The Decoder reports the model is aimed at everyday office tasks and automated software control, with examples including drafting proposals, interpreting technical drawings, and filling out forms.
SUPPORTEDChecked 2026-10-26 · 1 supporting
The Decoder reports improved price-to-performance and fewer hallucinations versus GPT-5.6 Terra, citing a drop in guessing on the AA-Omniscience benchmark from 88% to 38%.
SUPPORTEDChecked 2026-10-26 · 1 supporting
SOURCES & TIMELINE
2Bringing fast and capable computer use to everyone
OpenAI has released GPT-6.1 Terra, a new mid-range AI model at half the API price of its predecessor. OpenAI built the model for office tasks and automated software control. Benchmarks show a sharply better price-to-performance ratio and fewer hallucinations. OpenAI has released GPT-6.1 Terra, a model it says pushes the cost-to-performance boundary further than any of its predecessors. API prices are half those of…
OpenAI's Ted Sanders has categorized the new GPT-6.1 Terra and explained how the company names its language models. According to Sanders, model performance is multidimensional. A model that's great at math won't automatically be good at jokes. Still, most performance gaps come down to one variable: model size. Sanders says OpenAI's model names are difficult to simplify because model size, reasoning effort, and multi-…