Building infrastructure for advanced computing
The Computing Era
PerfXLab’s core team comes from the Chinese Academy of Sciences. We develop large language models and computing software stacks for LLMs, deep neural networks, and engineering computing.
A global accelerated-computing solutions provider, empowering chip, scientific computing, and HPC software companies.
Products

DF35 All-in-One Appliance
DF35 is a 4U rack-mount appliance with support for Qwen3.6-35B-A3B MoE and Qwen3.6-27B Dense models. PerfXLab’s PerfXLM inference framework optimizes the hardware and software together.
  • Performance breakthroughs: synergetic optimization of software and hardware
  • Planning for the future: ongoing technological evolution
Intelligent computing for industry digital transformation
Deep Fusion appliances deliver on-premises AI for businesses, public institutions, and teams with strict data security requirements. They come with a model development and management platform and support the full DeepSeek-R1/V3 671B and Qwen3-235B models.
Industry challenges
High technical threshold
Application complex
Cost pressure
Deployment and mobility challenges
Solution advantages
Smart Customisation Development
Data privacy security
Flexible expansion and eco-compatibility
Continuous optimization and technical support

Latest News

All News
Deep partnerships with 300+ computing companies
Helping our partners build advanced computing infrastructure
Join us in the computing era