Answer in brief
The new model is reaching Copilot's coding tools while OpenAI offers it through the API. Availability and price are documented; comparative task quality still needs local testing.
A release across coding surfaces
GitHub said on 29 September that GPT-6.1 Sol was generally available in Copilot, with a gradual rollout across supported editors, Copilot CLI, its coding agent, the Copilot app, github.com and mobile clients. OpenAI's API changelog separately lists the model's release for complex coding and professional work. Together, the notices establish distribution through two different routes. They do not imply that every account already sees the option in its model picker or that the same billing and controls apply in both routes.
What administrators and individual users actually control
GitHub names Pro+, Max, Business and Enterprise as eligible plans. Its announcement says Business and Enterprise administrators can manage the model through Copilot settings; under default enablement, a new model may be switched on unless an administrator has disabled the global default or this model. That setting matters because a team's approved models can differ from a developer's personal choices. A responsible rollout would check the organization's policy, the client where Sol appears, and which repositories or terminal tasks may use it before changing a production workflow.
Price is clearer than comparative performance
OpenAI lists standard API prices of $2 per million input tokens and $10 per million output tokens for prompts within the stated length band, and describes Sol as lower cost than Astra. GitHub says Copilot usage-based billing follows provider list pricing. GitHub also reports that its early tests completed tasks with fewer tokens and steps than earlier model families. Those are supplier observations, not a reproducible guarantee for an unfamiliar codebase; token savings can be lost if a task needs more retries, review or follow-up fixes.
The practical test is a completed change
A useful local comparison would give Sol and an existing approved model the same bounded coding task, then record whether each produced a passing change, how many tool steps and tokens it used, and what a reviewer had to repair. Security-sensitive work also needs normal code review and execution boundaries regardless of the selected model. At the 30 September evidence cutoff, the model's availability, controls and published prices are established; its value for a particular team remains an empirical question, not a result of the launch announcement.
Questions and answers
Who can select GPT-6.1 Sol in Copilot?
GitHub lists Pro+, Max, Business and Enterprise users, across its supported editors, CLI, coding agent, web and mobile surfaces. The rollout is gradual, and organizations may control access through model policy.
Does the lower API price guarantee cheaper coding work?
No. OpenAI's listed token price is one input. Total cost also depends on output volume, tool calls, retries and review time; GitHub says Copilot uses provider list pricing under usage-based billing.