Grok models with function calling, structured outputs, and live search grounding
Verified means we fetched it and it answered. Claimed means it is listed but has not been reached yet.
Claude models with tool use, the MCP connector, Agent Skills, and the Claude Agent SDK | 5/5 |
One OpenAI-compatible endpoint in front of hundreds of hosted models, with automatic fallback and per-key spend limits | 5/5 |
Unified model endpoint with failover, spend limits, and observability, wired into the AI SDK | 5/5 |
Gemini multimodal models via Google AI Studio and Vertex AI, with function calling and MCP support | 5/5 |
GPT-5 series models, embeddings, and the Responses API with native MCP and tool use | 5/5 |
Run models at the edge from Workers, with an OpenAI-compatible endpoint and AI Gateway | 5/5 |