Pointing Gemini CLI at custom models — your own endpoints, your own keys, your own cost — unlocks portability and cost control. The setup is straightforward; the discipline is what to check before you trust it.
The default Gemini CLI uses Google's models. Pointing it at custom models — your own endpoints, BYO keys, a different provider — is the move that gives you portability and cost control. The setup is easy; the trust comes from checking the right things.
The Opportunity: Portability and Cost
Custom models give you two things the default doesn't: portability (you're not locked to one provider's CLI) and cost control (you pick the model that fits the cost-quality trade). The setup is the unlock; the discipline is making sure the custom model behaves.
- Portability: not locked to one provider.
- Cost control: pick the model that fits the trade.
- BYO keys: the data flow is yours.
Sharpen the Saw: What to Check
Before you trust a custom-model setup, check the saw:
- **Endpoint and key.** The CLI points at your endpoint, with your key, and the connection works.
- **Tool compatibility.** The custom model supports the tools the CLI uses; not every model does.
- **Behavior.** The model behaves in the CLI's format — eval it on a small task before trusting it on real work.
- **Cost.** You know the per-request cost, and the budget holds.
The Setup Steps
- Configure the CLI for the custom endpoint and key.
- Confirm tool compatibility (run a small task that uses tools).
- Eval the behavior on a representative task.
- Set the cost expectations and monitor.
What to Refuse
- Trusting a custom model without a behavior eval.
- Assuming tool compatibility (not all models support all tools).
- Ignoring cost until the invoice.
- Setting it up without a fallback to the default.
The Fallback Discipline
Keep the default as a fallback. A custom model can break — the endpoint changes, the model degrades, the cost spikes — and the work shouldn't stop when it does. The setup that includes a fallback is the one that's safe to actually use.
Conclusion
Gemini CLI with custom models unlocks portability and cost control. Set it up, check the endpoint, the tools, the behavior, and the cost — and keep the default as a fallback. The trust comes from the checks, not the setup.
About FACTA
FACTA helps startups and growth-stage teams turn AI into production systems that keep running — not demos that impress once.
We design the architecture around the parts that actually break under real usage: tooling you own, credentials you control, failover, cost controls, observability. The boring infrastructure that keeps a system alive after launch.
Led by Matías Baglieri and Carolina Fogliato, we focus on one thing:
AI leadership that builds. Not just advises.
Tell us which custom model you'd point Gemini CLI at.
We'll tell you what to check before you trust it. See how to choose an AI model for the model side.
Explore AI Automation