CONFIRMEDDate confidence: CONFIRMEDStated as firmHow firmly the source words the date: confirmed, scheduled, estimated, delayed or unverified.Show all CONFIRMED pins90% CONFIDENCEOverall confidence: 90%How well the pin's source and references back up its dates.Weighted average of how firmly 6 references, the source included, support the pin's start and end times; a reference counts half as much for every 180 days older than the newestShow all pins at 75% confidence or better
Stated: OpenAI[1][2] says “Starting today”; independent launch-day articles from Artificial Analysis[3] and Plotly[4] are explicitly dated October 26, 2026.- Start
- Oct 26, 202690% CONFIDENCEfrom the source
GPT-6.1 Terra Released
Main competitors
- Claude Sonnet 5.5 · Anthropic
Competes for general-purpose coding and computer-use workloads; OpenAI explicitly includes Sonnet 5.5 in its launch comparison charts.
Source - Claude Opus 5.5 · Anthropic
Higher-cost alternative for demanding agent tasks, directly compared in the launch’s task-cost and scientific-terminal evaluations.
Source
OpenAI releases its lower-cost Terra model for visual reasoning, computer use and writing. Independent launch tests show useful efficiency gains alongside task-specific accuracy limits.
- OpenAI released GPT-6.1 Terra on October 26, 2026 for the API and paying ChatGPT Work and Codex plans. It follows July’s GPT-5.6 Terra and targets everyday computer-based work at lower cost than the larger Sol and Astra models.[1]
- Independent launch-day assessments show the trade-off depends on the task: Artificial Analysis places its maximum-effort intelligence score between the current Sol and Luna models, while its lower effort settings lead on cost efficiency. Its benchmark version is v4.3.2, rather than an older scale.[3]
- Plotly’s own synthetic data-analysis exam found 40 of 43 correct at default reasoning, costing $1.31 for the run. Its three misses included hidden patterns and a causal explanation; this is a particular workload, not a universal accuracy guarantee.[4]
- DualEntry’s separate 101-task accounting test found 78.2% accuracy and only 65.3% consistent success across three runs. The accounting-software vendor highlights weaknesses in payables, receivables and month-end close, tempering the broad launch claims.[5]
Notable features
- The API accepts text and images and produces text, with a 1,050,000-token context window and up to 128,000 output tokens. Audio and native video input are unsupported; operating a video editor is distinct from accepting video as a model input.[2]
- Standard short-context rates are $1 input, $0.05 cached input and $5 output per million tokens. That is $2 per million at a 3:1 input/output mix, before cache, tools or long-context adjustments. API usage is metered, so no single consumer purchase price is assigned.[2]
- OpenAI reports 50.3% on Terminal-Bench 4.0 at maximum effort, versus 21.5% for its predecessor. The maker’s internal factuality test reports 16.7% errors at low effort and 4.3% at maximum; these internal rates are distinct from independent accounting or analytics accuracy.[1]
Buy onChatGPT