GitHub Copilot Retires Six Models September 1: Budget the Migration Now

GitHub says Gemini 3.1 Pro, Claude Opus 4.5, Opus 4.6, Sonnet 4.5, most Sonnet 4.6 access, and Raptor Mini leave Copilot on September 1. Suggested replacements exist, but model substitution changes output, token use, policy settings, and review behavior.

The factual baseline comes from GitHub's Copilot deprecation notice. Product behavior and prices can change, so confirm the current page before making a purchase or locking an annual budget.

The budgeting mistake is to look at one visible rate and treat it as the whole system. Coding work crosses planning, context retrieval, tool calls, file edits, execution, validation, repair, and review. Every stage can add tokens, runtime, or people. A useful model preserves those layers and then rolls them into cost per accepted result.

What Matters for the Cost Model

• GitHub recommends Gemini 3.6 Flash in place of Gemini 3.1 Pro.
• Claude Opus users can move to Opus 4.7, Opus 4.8, or Opus 5, while Sonnet users are directed to Sonnet 5.
• Raptor Mini users are pointed to MAI-Code-1-Flash in the published notice.
• Enterprise administrators may need to enable replacement models through model policies before users can select them.

These facts change the shape of spend, but they do not determine whether the product is economical. A higher-priced run can be the cheaper outcome if it merges on the first attempt and needs little review. A low-cost run becomes expensive when it wakes repeatedly, occupies compute while idle, or produces a patch that experts must rewrite.

Build the Equation Before the Forecast

Migration cost equals evaluation tasks multiplied by candidate models and run cost, plus prompt updates, policy work, reviewer time, and regression repair. Testing 40 tasks on two replacements at $0.35 per run costs only $28 in model usage, but eight reviewer hours at $100 adds $800.

Keep cached input, uncached input, and output separate because their rates can differ by an order of magnitude. Add tool fees, search calls, image or screenshot processing, and regional premiums only when the workflow actually uses them. For runtime, model billing minimums and rounding explicitly. A spreadsheet that assumes perfect per-second billing can understate a block-priced service.

A Worked Monthly Example

A 60-developer team can start with a stratified 30-task suite: completion, refactor, test generation, debugging, and security review. Run the incumbent while still available, then compare candidates on acceptance rate, output tokens, latency, and reviewer corrections. A two-point acceptance drop may outweigh a lower token rate.

Run a sensitivity range around the estimate. Use a low, expected, and high case for acceptance rate, output length, retry count, and runtime. Those variables usually move the answer more than a tiny difference in input-token price. The high case should represent a difficult but plausible sprint, not an impossible catastrophe.

Controls That Preserve the Savings

• Inventory pinned and default models. Make this visible in the run record so the team can connect the control to dollars and accepted output.
• Enable replacements before the deadline. Make this visible in the run record so the team can connect the control to dollars and accepted output.
• Replay representative prompts with fixed tests. Make this visible in the run record so the team can connect the control to dollars and accepted output.
• Communicate changed strengths and fallback behavior. Make this visible in the run record so the team can connect the control to dollars and accepted output.

Give every automated run a durable identifier that follows retries, branches, builds, and review. Without that join key, invoices remain disconnected from engineering outcomes. Aggregate by task class, repository, team, model, and trigger so a change in workload mix does not masquerade as a price change.

The Failure Mode to Watch

Silent fallback is worse than an explicit migration because teams may attribute changed behavior to prompts or repositories. Log model identity where the product exposes it, update internal documentation, and watch for extensions or automations that assume a retired model name.

Set a stop condition before the run begins. Repeated failure on the same test, no material repository change, a spend ceiling, or a deadline can all trigger escalation. Stopping is not failure; it protects the budget and gives a human the evidence needed to change the approach instead of financing another identical loop.

How to Decide

Treat deprecation as a small production migration. Use the remaining window to establish a baseline, select the least disruptive replacement, and keep a second fallback. The token bill for evaluation is usually trivial compared with confused developers and emergency prompt rewrites after cutoff.

Start with a bounded pilot and preserve a control group using the current process. Track model dollars, runtime dollars, tool charges, reviewer minutes, acceptance, cycle time, and post-merge defects. After enough comparable tasks, calculate total cost per accepted result. Use the AI Cost Calculator for the token component, then add the operational layers the API invoice does not show.