SpaceXAI introduced Grok 4.5, its latest AI model focused on coding, engineering, and knowledge tasks. With significant improvements in training techniques and data quality, it outperforms previous models in efficiency and intelligent reasoning.
SpaceXAI released Grok 4.5, its latest and most advanced AI model. This model is specifically built to excel in coding, agentic tasks, and various knowledge work domains. It was developed alongside Cursor, ensuring top performance standards.
Grok 4.5 was trained on extensive datasets covering coding, science, engineering, and math. The training utilized tens of thousands of NVIDIA GB300 GPUs, with new methods focused on data filtering and curation to raise overall data quality. As a result, Grok 4.5 demonstrates superior performance over leading models in real engineering tasks.
The model uses reinforcement learning that encompasses hundreds of thousands of tasks in multi-step software engineering and other technical efforts. This approach supports asynchronous training, allowing agentic rollouts to run for extended hours across multiple GPUs, leading to an increase in intelligent reasoning capabilities.
Grok 4.5 is highly proficient in coding across various languages like Rust and C/C++. It can handle complex tasks such as end-to-end application development efficiently. Additionally, Grok Build, which now uses Grok 4.5 as its default model, supports creating intricate Excel models and enhancing productivity in software like PowerPoint and Word.
The model achieves a delivery speed of 80 transactions per second (TPS) and boasts twice the token efficiency compared to previous leading models. This improvement enables faster and cost-effective processing capabilities for users, making Grok 4.5 an attractive option for businesses that rely on AI in technical tasks.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
SpaceXAI introduced Grok 4.5, its latest AI model focused on coding, engineering, and knowledge tasks. With significant improvements in training techniques and data quality, it outperforms previous models in efficiency and intelligent reasoning.