microsoft/NextCoder-14B

microsoft/NextCoder-14B is a 14.7 billion parameter causal language model developed by Microsoft, based on the Qwen2.5-Coder Instruct variants. It is specifically fine-tuned for robust code editing tasks using the novel Selective Knowledge Transfer (SeleKT) methodology. This model demonstrates significant improvements in code editing performance, particularly on complex benchmarks like Aider-Polyglot, while maintaining generalization capabilities and supporting a long context of up to 32K tokens.