CodeQwen1.5 is based on Qwen1.5. It is trained on 3 trillion tokens of code data. Its major features include:

• Strong code generation capabilities and competitive performance across a series of benchmarks
• Support for long context understanding and generation with a maximum context length of 64K tokens
• Support for 92 coding languages
• Excellent performance in Text-to-SQL, fixing bugs and other coding use cases.

References

Blog Post

GitHub

HuggingFace