StableCode is an LLM generative AI product for coding developed by Stability AI. It aims to assist programmers in their daily work and serve as a learning tool for new developers. The tool offers three different models to enhance coding efficiency. The base model is trained on a diverse set of programming languages from the stack-dataset, including popular languages like Python, Go, Java, JavaScript, C, Markdown, and C++. It has been further trained on 560B tokens of code. The instruction model is specifically tuned to solve complex programming tasks and is trained on around 120,000 code instruction/response pairs. StableCode's long-context window model allows for single and multiple-line autocomplete suggestions, making it an ideal assistant for reviewing or editing large amounts of code simultaneously. Compared to previous open models, StableCode can handle 2-4 times more code at once, equivalent to editing up to five average-sized Python files. This feature makes it an excellent learning tool for beginners who want to tackle more significant coding challenges. Stability AI aims to make technology more accessible, and StableCode is a significant step in realizing this vision. The tool empowers people of all backgrounds to create code to solve everyday problems using AI. It also seeks to provide fairer access to technology worldwide. StableCode is designed to help the next generation of software developers learn to code and contribute to a more inclusive tech ecosystem.
