Skip to content
NewsProduct launch

GitHub Copilot adds OpenAI's GPT-6.1 Terra as a cheaper everyday coding model

· by

The short version

GPT-6.1 Terra reached general availability inside GitHub Copilot on October 26, 2026, priced at half of GPT-6.1 Sol's rate while more than doubling the prior Terra tier's score on OpenAI's own terminal-coding benchmark.

GitHub Copilot adds OpenAI's GPT-6.1 Terra as a cheaper everyday coding model

OpenAI's GPT-6.1 Terra reached general availability inside GitHub Copilot on October 26, 2026, priced at half of what GPT-6.1 Sol and the prior GPT-5.6 Terra charge, while more than doubling Terra's score on OpenAI's own command-line coding benchmark, per GitHub's changelog and OpenAI's launch post.

What happened

GitHub made Terra selectable in the Copilot model picker for Pro, Pro+, Max, Business, and Enterprise plans, across VS Code, Visual Studio, Copilot CLI, the cloud agent, the Copilot app, github.com, GitHub Mobile, JetBrains IDEs, Xcode, and Eclipse, GitHub's changelog says. Rollout is gradual, and Business and Enterprise admins control access through Copilot's model policy; new models are enabled by default unless an admin has turned off global default enablement or blocked this one specifically. Usage is billed at provider list pricing under usage-based billing rather than a fixed per-request multiplier.

OpenAI's own announcement, published the same day, sets GPT-6.1 Terra's standard API price at $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens, per OpenAI. That is 50% below both GPT-6.1 Sol's standard rate and the prior GPT-5.6 Terra's rate. On Terminal-Bench 4.0, a test of command-line coding tasks, GPT-6.1 Terra solves 50.3% of tasks at maximum reasoning effort, more than double GPT-5.6 Terra's published score of 21.5%, OpenAI reported. GitHub's own changelog echoed that jump in plainer terms, saying early testing showed Terra handling noticeably more terminal and C# tasks than GPT-5.6 Terra.

Terra narrows the gap to Sol, but OpenAI says it still has a ceiling

OpenAI's launch post lists other gains alongside the coding score. On its internal hallucination-prone-prompt benchmark, Terra's error rate at low reasoning effort fell from 53.6% with GPT-5.6 Terra to 16.7%, and at maximum effort to 4.3%, nearly matching GPT-6 Astra and GPT-6.1 Sol. On a computer-use benchmark called Agents' Last Exam, Terra at medium effort scored 15 points higher than GPT-5.6 Terra at about a third of the cost, and matched GPT-6 Astra's result at one-tenth Astra's price per token.

That combination changes the default calculus in Copilot's model picker. A team picking a model for everyday agentic coding or terminal work no longer has to trade much capability for Terra's lower price; the quality gap that justified paying for Sol on routine work has mostly closed. OpenAI itself draws the line on where it has not closed: "For the most complex coding tasks, GPT-6.1 Sol and GPT-6 Astra remain the better choice," the company wrote. Cost-conscious teams can move their day-to-day default to Terra and reserve Sol or Astra for the hardest problems, rather than running everything through the pricier tier out of habit.

What to watch next

Watch how Terra's $1 input and $5 output per-million pricing holds up against comparable Claude and Gemini tiers once the rollout finishes across every Copilot surface GitHub listed. Also watch for an independent benchmark run, since every performance number in OpenAI's launch post, including the Terminal-Bench 4.0 and Agents' Last Exam scores, comes from OpenAI's own evaluation suite rather than a third party.

Sources