InferenceX

InferenceX™ is an open-source inference performance research platform dedicated to continually analyzing and benchmarking the world's most popular open-source inference frameworks. It tracks real-world model inference performance, providing near real-time progress indicators. The platform supports various hardware and software stacks, including NVIDIA Blackwell, AMD GB300, and more, offering a publicly accessible, free live dashboard. Its core aim is to capture the evolving performance gains from AI software stacks like SGLang and vLLM.

• Open-source inference performance benchmarking platform
• Real-time tracking of model inference performance
• Supports multiple mainstream hardware and AI frameworks
• Provides public performance dashboards
• Focuses on capturing continuous AI software performance optimizations