GitHub Copilot Adds GPT-5.6 Sol, Terra, and Luna: Which One to Pick

GitHub just widened the model picker inside Copilot again, and this time the split isn't cosmetic. As of July 9, three OpenAI GPT-5.6 variants - Sol, Terra, and Luna - are live across the Copilot lineup, and each one is built for a different shape of coding work rather than a different price point on the same model.

If you've been defaulting to whatever Copilot suggests, this is a good moment to stop doing that.

What actually shipped

The three models split cleanly by job, not by version number. Sol is described as the highest-reasoning-ceiling option in the family, aimed at complex reasoning over large codebases and long-running agentic work - the kind of task where you kick off a multi-file refactor or a dependency migration and walk away for twenty minutes. Terra is the balanced default, tuned for the everyday back-and-forth of interactive and agentic coding most people actually do most of the day. Luna is the lightweight, cost-efficient tier - the cheapest of the three - built for small, fast tasks where you don't want to burn a premium request on a one-line fix.

Here's the part that will actually change your workflow: access isn't uniform. Sol requires Copilot Pro+, Max, Business, or Enterprise - plain Pro subscribers don't get it. Terra and Luna, by contrast, are available across the full paid range, from base Pro up through Enterprise.

Where it runs, and what it costs

All three models reach Copilot's full surface area: VS Code, Visual Studio, the Copilot CLI, the GitHub Copilot cloud agent, the GitHub Copilot mobile app, github.com itself, GitHub Mobile on iOS and Android, JetBrains IDEs, Xcode, and Eclipse. That's a wider simultaneous rollout than some past Copilot model additions, which sometimes staggered by IDE for weeks.

Billing runs through Usage Based Billing at provider list pricing, which means Sol - the heavier-reasoning model - will cost more per call than Terra or Luna in practice, even though GitHub hasn't published an exact per-model multiplier alongside the announcement. If your team is on a metered plan rather than flat-rate seats, that's worth watching closely for the first billing cycle after rollout. The gap is not subtle: at OpenAI's published list prices - $6 per million output tokens for Luna, $30 for Sol - an agent pipeline pushing 100M output tokens a month costs about $600 on Luna and $3,000 on Sol. Our GPT-5.6 tier pricing breakdown runs the full math across all three tiers. Defaulting every agentic task to Sol moves that needle five times faster than defaulting to Luna.

According to GitHub's supported-models reference, all three ship at General Availability status - not preview - and each carries a 1 million token context window with configurable reasoning levels. That context size matters more for Sol's stated use case (large-codebase reasoning) than for Luna, where most tasks won't come close to using it.

The Enterprise catch

If you're on Business or Enterprise, don't expect GPT-5.6 to show up automatically for your team. The policy is off by default, and an admin has to explicitly enable it before any seat can select Sol, Terra, or Luna. This is consistent with how GitHub has gated other recent model additions to Copilot for organizational accounts, and it's an easy step to miss if you're not the one managing the tenant.

Solo Pro and Pro+ subscribers don't have this gate - Terra and Luna are simply available in the model picker now, and Pro+ users additionally see Sol.

Where this leaves your Copilot setup

Pick by task shape, not habit. If you're running a long agentic session - a big refactor, a multi-repo migration, anything where the agent will be making dozens of sequential decisions - Sol is the one to reach for, assuming your plan includes it. For the routine work that fills most of a coding day - writing a function, fixing a test, answering a quick question about the codebase - Terra is the sensible default; it's built for exactly that and it's available on every paid tier. Luna is for the stuff you'd almost do yourself: a rename, a one-line type fix, a quick explain-this-error. Using Luna there instead of Terra or Sol is the cheapest lever you have under Usage Based Billing.

For teams already comparing coding assistants, see ToolPick's GitHub Copilot pricing breakdown for how Sol's Pro+/Max/Business/Enterprise gating lines up against current plan costs, and the Claude Code vs. Cursor vs. GitHub Copilot comparison for how this model split affects Copilot's standing against its two closest rivals.

One footnote worth flagging for teams tracking the OpenAI-Microsoft relationship: OpenAI remains an independent Public Benefit Corporation, with Microsoft holding a minority stake rather than ownership. GitHub itself is a Microsoft product as before - nothing about that structure changed with this rollout.