Using GitHub Copilot CLI with Venice.ai
I've seen a lot of complaints about the new token based billing for GitHub Copilot. The previous pricing plans were just too good to be true. We're in a new reality now, so what can you do?
I'll tell you about my setup:
- GitHub Copilot CLI
- Custom LLM Provider
- Cheaper models
I'm using GitHub Copilot CLI with venice.ai as custom LLM model provider. But you can also use OpenRouter, or other OpenAI compatible API endpoints.
Models
Kimi 2.6 has been working great for me on my C# projects and is way more cheaper than Claude Sonnet models op GPT 5. Currently I'm testing the MiniMax M3 model, which is even cheaper! Only for really difficult tasks, where I'm not getting the results I want, I switch to a more expensive model.
Pricing
| Model | Input | Output |
|---|---|---|
| MiniMax M2 | $0.300 | $1.20 |
| Kimi K2.6 | $0.850 | $4.66 |
| Claude Opus 4.8 | $6.00 | $30.00 |
| GPT5.5 | $6.25 | $37.50 |
Conclusion: use the cheaper models as much as you can. Only when they do not deliver the results you want, switch to more expensive models.
Bring Your Own Model setup
First, install the GitHub Copilot CLI Start PowerShell and paste this:
$env:COPILOT_PROVIDER_BASE_URL = "https://api.venice.ai/api/v1"
$env:COPILOT_PROVIDER_API_KEY = "TODO_VENICE_INFERENCE_KEY"
$env:COPILOT_MODEL = "kimi-k2-6"
copilot
This will start copilot, configured for Venice and using the kimi-k2-6 model. For MiniMax M3, change it to:
$env:COPILOT_MODEL = "minimax-m3"
Here is the documentation about configuring your own models: https://docs.github.com/en/copilot/how-tos/copilot-cli/customize-copilot/use-byok-models#connecting-to-an-openai-compatible-endpoint