Credit tips
Spend credits on purpose.
Separate tabs for Auto, Chat, agents, what you paste, and which model to open. Use the planner on Credits to pick models for your leftover this month.
Reviewed
GitHub Auto
Choose the right Auto tier
Auto considers task complexity, model availability, your plan, and organization policies. A discount reduces the selected model's cost; it does not make every request equally cheap.
01
Turn Auto on for everyday paid Copilot
Paid Copilot plans receive a 10% discount on model costs with Auto in Chat, CLI, the Copilot app, and cloud agent. Treat any additional promotion as temporary and check your billing page before budgeting around it.
02
Efficiency, Balance, or Intelligence
Where Auto tiers are available, choose Efficiency for straightforward work, Balance for everyday tasks, and Intelligence for complex work. Tiers influence routing preferences; they are not fixed model lists or hard spending caps.
03
Avoid unnecessary manual switches
Manual model changes can reduce cache reuse. Auto routes around natural cache boundaries and substantial changes in task complexity. Keep a stable brief; switch when the task benefits, not just to chase a different name.
04
Check which model actually answered
Hover over a response in Copilot Chat or the Copilot app to inspect the model. CLI displays it in the terminal. If usage rises, compare task size, reasoning effort, and selected model before changing your whole workflow.
Based on official GitHub and provider documentation. API rates, Copilot availability, promotions, and plan billing can differ.
- GitHub: models and billing
- GitHub: Auto model selection
- GitHub: availability and retirements
- Anthropic models and pricing
- OpenAI models and pricing
- Google models
- Google pricing
- DeepSeek models and pricing
- xAI models and pricing
- Kimi models and pricing
- Mistral Medium 3.5
- Mistral Small 4
- Qwen models
- Qwen international pricing