StarCoder
StarCoder is a family of open-access code large language models (LLMs) from BigCode, an open scientific collaboration between Hugging Face and ServiceNow, trained on permissively licensed repositories to complete, generate, and explain source code. The family spans two generations:
The original 15.5-billion-parameter release from 2023 is Python-specialized. BigCode built it by fine-tuning the multilingual StarCoderBase checkpoint on 35 billion Python tokens, and a smaller 164-million-parameter variant, tiny_starcoder_py, trains on Python alone.
StarCoder2 followed in 2024, developed with NVIDIA and Software Heritage in 3B, 7B, and 15B sizes. All three train on The Stack v2 and widen the context window to 16,384 tokens, though only the 15B model spans the dataset’s full 600-plus languages. The base models support fill-in-the-middle completion, which fills a gap between a given prefix and suffix.
Neither generation is instruction-tuned. BigCode’s model cards warn that plain instructions such as “write a function that computes the square root” don’t work well, so chat-style use calls for a tuned derivative like starcoder2-15b-instruct-v0.1.
The weights ship on Hugging Face under the BigCode OpenRAIL-M license, which permits commercial use with responsible-use restrictions. Python developers run them locally through the transformers package or a runner like Ollama.
No newer generation has shipped since StarCoder2, and BigCode’s later work centers on datasets and benchmarks. Open-weight peers include Code Llama and CodeGeeX, along with later families like DeepSeek-Coder and Qwen3-Coder that continue to ship new generations.
Official website: BigCode
Related Resources
Tutorial
Hugging Face Transformers: Leverage Open-Source AI in Python
As the AI boom continues, the Hugging Face platform stands out as the leading open-source model hub. In this tutorial, you'll get hands-on experience with Hugging Face and the Transformers library in Python.
For additional information on related topics, take a look at the following resources:
- How to Use Ollama to Run Large Language Models Locally (Tutorial)
- How to Integrate Local LLMs With Ollama and Python (Tutorial)
- GitHub Copilot: Fly With Python at the Speed of Thought (Tutorial)
- How to Use Google's Gemini CLI for AI Code Assistance (Tutorial)
- How to Use the OpenRouter API to Access Multiple AI Models via Python (Tutorial)
- Hugging Face Transformers (Quiz)
- How to Use Ollama to Run Large Language Models Locally (Quiz)
- How to Get Started With Ollama (Course)
- How to Integrate Local LLMs With Ollama and Python (Quiz)
- Getting Started With Google Gemini CLI (Course)
- How to Use Google's Gemini CLI for AI Code Assistance (Quiz)
- Accessing Multiple AI Models With the OpenRouter API (Course)
- How to Use the OpenRouter API to Access Multiple AI Models via Python (Quiz)
By Martin Breuss • Updated Aug. 25, 2026
