This paper introduces bGPT, a model designed for binary data processing and digital world modelling through next byte prediction. Unlike traditional language models that focus on text, bGPT extends deep learning to binary data, enabling it to simulate the digital world by interpreting and manipulating binary data. The model is capable of predicting, simulating, and diagnosing algorithm or hardware behavior, and has achieved a low error rate of 0.0011 bits per byte in converting ABC notation to MIDI format. It also demonstrates exceptional capabilities in simulating CPU behavior, with an accuracy exceeding 99.99% in executing various operations. The paper discusses the limitations of traditional deep learning models that primarily operate within the realm of media data, overlooking the omnipresent native binary data in the digital world. bGPT addresses this by directly interpreting and manipulating binary data, enabling a more intrinsic and holistic understanding of the digital world. It offers two main advantages: 1) interpreting digital systems by learning the patterns of digital systems, enabling it to predict, simulate, and diagnose algorithm or hardware behavior, and 2) unified modelling by integrating various data types into a single framework, treating everything as a byte sequence. The paper presents the methodology of bGPT, including its hierarchical Transformer framework, which segments byte sequences into patches to manage computational efficiency. The model is trained for generative modelling through next byte prediction and classification tasks. The training objectives include generative modelling and classification, with the model being evaluated on various tasks, including digital media processing and algorithm and hardware simulation. The paper also discusses the applications of bGPT, including digital media processing and algorithm and hardware simulation. It demonstrates the model's capabilities in data conversion and CPU state modelling, showing its potential for simulating various algorithms and hardware. The experiments show that bGPT performs well in these tasks, with the model achieving near-perfect performance in data conversion and CPU state modelling. The paper concludes that bGPT demonstrates strong scalability on native binary data with emergent abilities in data conversion and CPU state modelling, which illuminate its potent capabilities in algorithm and hardware simulation tasks. However, the paper also highlights the need for further research to improve the computational cost and scalability of byte models. The paper also discusses the ethical implications and potential impact of bGPT on societal norms and legal frameworks.This paper introduces bGPT, a model designed for binary data processing and digital world modelling through next byte prediction. Unlike traditional language models that focus on text, bGPT extends deep learning to binary data, enabling it to simulate the digital world by interpreting and manipulating binary data. The model is capable of predicting, simulating, and diagnosing algorithm or hardware behavior, and has achieved a low error rate of 0.0011 bits per byte in converting ABC notation to MIDI format. It also demonstrates exceptional capabilities in simulating CPU behavior, with an accuracy exceeding 99.99% in executing various operations. The paper discusses the limitations of traditional deep learning models that primarily operate within the realm of media data, overlooking the omnipresent native binary data in the digital world. bGPT addresses this by directly interpreting and manipulating binary data, enabling a more intrinsic and holistic understanding of the digital world. It offers two main advantages: 1) interpreting digital systems by learning the patterns of digital systems, enabling it to predict, simulate, and diagnose algorithm or hardware behavior, and 2) unified modelling by integrating various data types into a single framework, treating everything as a byte sequence. The paper presents the methodology of bGPT, including its hierarchical Transformer framework, which segments byte sequences into patches to manage computational efficiency. The model is trained for generative modelling through next byte prediction and classification tasks. The training objectives include generative modelling and classification, with the model being evaluated on various tasks, including digital media processing and algorithm and hardware simulation. The paper also discusses the applications of bGPT, including digital media processing and algorithm and hardware simulation. It demonstrates the model's capabilities in data conversion and CPU state modelling, showing its potential for simulating various algorithms and hardware. The experiments show that bGPT performs well in these tasks, with the model achieving near-perfect performance in data conversion and CPU state modelling. The paper concludes that bGPT demonstrates strong scalability on native binary data with emergent abilities in data conversion and CPU state modelling, which illuminate its potent capabilities in algorithm and hardware simulation tasks. However, the paper also highlights the need for further research to improve the computational cost and scalability of byte models. The paper also discusses the ethical implications and potential impact of bGPT on societal norms and legal frameworks.