GPT-6.1 Terra is now available in two environments already used by many technology teams: GitHub Copilot and Amazon Bedrock. The October 26 announcements bring the model to both coding assistance and applications running through AWS infrastructure.
The starting point is where the work happens. Selecting a model inside a coding assistant and integrating it into a service involve different configuration, billing and access-control decisions.
In Copilot, start with the model picker
GitHub is making Terra available to Pro, Pro+, Max, Business and Enterprise subscribers through a gradual rollout. Supported environments include VS Code, Visual Studio, CLI, the cloud agent, the Copilot app, github.com, mobile devices and IDEs such as JetBrains, Xcode and Eclipse.
GitHub describes its focus as everyday agentic coding and terminal work, particularly well-scoped tasks. It reports improvements over GPT-5.6 Terra in early terminal and C# testing, but the announcement does not provide a detailed scorecard to quantify the gain.
Business and Enterprise access depends on model policies. The global default can enable new models automatically unless an administrator has disabled that setting or blocked Terra specifically. The model is billed at provider list pricing under usage-based billing.
In Bedrock, check the integration
AWS announced general availability through APIs, with the option to configure ChatGPT Work and Codex to use the model through Bedrock. The announcement lists text and image inputs and a context window of up to 1 million tokens.
The AWS technical documentation for Terra distinguishes between the bedrock-mantle and bedrock-runtime endpoints. Features, routing and regional availability differ. Existing applications therefore need a compatibility check before switching, especially if they depend on server-side tools or data staying within a specific region.
Lower token pricing also needs to be considered alongside that configuration. The documentation separates short and long context pricing and different inference options. A lower unit price can help, but task cost also depends on the tokens and attempts needed to finish the work.
Compare it on your own tasks
If your team already uses Copilot, familiar tasks provide a useful basis for comparing changes, review time and consumption. For Bedrock, include latency, tool calls and failure behavior.
A small fix and an open-ended code investigation may benefit from different models. Terra expands that choice; the useful comparison puts the delivered result and the review effort in the same calculation.
Official sources in English. Announcements dated October 26, 2026; documentation checked on October 27, 2026.