Blog · 2026-06-07
Building anything serious with LLMs usually means hedging across providers: Claude for long-context reasoning, GPT for tool use, Gemini for cheap multimodal. Officially that’s three accounts, three keys, three billing setups, and three regional-access headaches.
API Models exposes all three behind one OpenAI-compatible endpoint. Point any OpenAI SDK at https://apimodels.app/api/v1, and switch models by changing the "model" field — claude-opus-4-8, gpt-5.1, gemini-3-pro-preview. One key, one balance, one invoice.
This makes routing and fallback trivial: try the cheapest model first, fall back to a stronger one on low confidence, all without re-authenticating or re-instrumenting. And it’s reachable from mainland China without a VPN.
Yes — /v1/chat/completions is OpenAI-compatible and /v1/messages is Anthropic-compatible. Existing SDKs work unchanged.