The 'Qimeng' system, a collaborative effort between the National Key Laboratory of Processor Chips at the Institute of Computing Technology and the Institute of Software, Chinese Academy of Sciences, stands as the world's inaugural fully automated design system for processor chip hardware and software, leveraging artificial intelligence technology. It seamlessly automates the entire design process, spanning from chip hardware architecture to fundamental software. Under the guidance of three academicians, this system employs large-scale models, intelligent agents, and Boolean logic generation technology to overcome the limitations of traditional chip design, which is often hindered by manual iteration, time-consuming processes, and significant resource demands.

Its core accomplishments encompass:

1) Chip Design Automation: The 'Qimeng-1' successfully accomplished the front-end design of a 32-bit RISC-V CPU in just 5 hours, boasting a scale of 4 million logic gates. Its performance is on par with the Intel 486, and it has successfully completed tape-out. The 'Qimeng-2', recognized as the first fully automated superscalar processor core, expanded its scale to 17 million logic gates, matching the performance of the ARM Cortex A53.

2) Basic Software Optimization: The automatic configuration of the operating system kernel has enhanced performance by 25.6% compared to manual optimization. Meanwhile, cross-platform tensor program translation has achieved double the performance of vendor-optimized operator libraries. Matrix multiplication operators have witnessed performance improvements of 110% and 15% on RISC-V CPUs and NVIDIA GPUs, respectively.

3) Methodological Innovation: It has introduced a verification-centric intelligent design methodology, generating circuit logic through binary guessing graph technology combined with Monte Carlo sampling and gate-level dependency analysis. Furthermore, it has constructed a three-tier architecture encompassing a large model in the processor chip domain, chip/software intelligent agents, and application levels. This architecture supports a self-evolving design paradigm that harmonizes top-down and bottom-up approaches.

This system has been honored with the Leading Science and Technology Award at the 2025 World Internet Conference. Additionally, its related achievements have been accepted by prestigious international conferences such as IJCAI 2024, marking a significant leap forward for China in the realm of AI-driven automatic chip design.