Paid services on Arbitrum matching “code”

OpenAI-compatible LLM inference, paid per call. POST a standard chat completion body with an x402 payment and the response is the completion itself; the API key and remaining balance come back in headers, so the next call can skip payment. Serves MiniMax language models: MiniMax-M3 for long-context reasoning with a 1,048,576-token window, and MiniMax-M2.7 for general chat, agent loops and coding.