InferX Beta — Serverless GPU Inference Platform, Built for Agent-Native Workloads

Moonlight-16B-A3B is a high-efficiency Mixture-of-Experts (MoE) language model released in February 2025 by Moonshot AI (the creators of Kimi). It was designed to push the "Pareto frontier"—delivering the reasoning power of much larger models while maintaining the inference speed and VRAM footprint of a small model