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

ykarout/Qwen3.5-9B-NVFP4 is a 9B Qwen3.5 model optimized using NVIDIA NVFP4 4-bit floating-point quantization, significantly reducing GPU memory usage and improving throughput while maintaining strong reasoning, coding, and chat capabilities, making it ideal for cost-efficient deployments and high-concurrency inference. NVFP4 quantization can reduce memory usage by roughly 4× compared to BF16 while preserving useful numerical properties for LLM inference.