The platforms below offer free models or free credits for new users. Click “Get API Key” to sign up and grab your key, then paste it into ModelHub to start using it. Quota policies change often — always check each platform's console for actual availability.
After you get a key: Open the ModelHub manager (start.html) → “Configure API Keys” card, and paste the key under the ModelHub provider labeled on each card; then in the “Models” list, pick that platform's free model alias (listed on the card).
Protocol compatibility: Claude Code users don't need to worry about protocols — the ModelHub gateway auto-converts the Anthropic protocol to these platforms' OpenAI protocol (including tool_use), works out of the box. Codex users can also bypass the gateway and point base_url to the platform's OpenAI-compatible endpoint directly in ~/.codex/config.toml using model_providers.
Quota exhausted / model retired: If a platform's free quota runs out or a model is retired, manually switch to another free model or your paid key in the ModelHub model list (no auto-fallback in the current MVP stage). Free models generally have concurrency / RPM limits (e.g. Zhipu 30 concurrent, SiliconFlow 1000 RPM); for high-concurrency tasks use a paid tier or off-peak hours.