copilot's model picker now comes with a credit meter
github added gemini 3.6 flash to copilot on july 21, one day after shipping clearer ai-credit usage and cost-center controls. the model menu is getting richer, but the meter is getting closer too.
github's Copilot model picker picked up another serious coding model this week: Gemini 3.6 Flash started rolling out on july 21, 2026 for Copilot Pro, Pro+, Max, Business, and Enterprise users. github says the model is aimed at web and app development, coding, and longer-horizon agentic tasks, with configurable reasoning effort and parallel tool use.
the timing matters. one day earlier, github also shipped two billing controls around Copilot AI credits: Business and Enterprise users can now see the number of AI credits used in the current billing cycle, and enterprise admins can manage AI-credit pools for cost centers directly in the billing UI.
choice is becoming an ops problem
copilot's model menu is no longer a simple "best model" dropdown. in july alone, github announced Gemini 3.6 Flash, OpenAI GPT-5.6 Sol/Terra/Luna, and Kimi K2.7 Code as selectable coding options. each one comes with its own availability rules, admin policy switches, and usage-based billing note.
that is good for developers who know what they are doing. it is also the exact moment a team needs norms. who gets to switch models? when should someone use the cheapest fast model instead of the long-running agent model? does the repo have a default in its instructions, or does everyone pick by vibes at 2am?
the boring policy is the feature
for individual users, the practical change is simple: check your AI credits before and after a long agent session until you understand your burn rate. for teams, the better move is to pair model access with lightweight policy: default model by repo, escalation rules for expensive reasoning, and a monthly review of credit usage by team or cost center.
the model picker is turning into a tiny procurement system inside the editor. treat it that way. put the rules near the code, and make the cost visible before the invoice does.
if you are setting those rules from scratch, start with the four prompts before any feature and add a fifth line: which model is allowed to spend real money on this task?