DeepSeek Coder 1.3B

About

DeepSeek Coder 1.3B is a robust language model designed for coding tasks, featuring 1.3 billion parameters. It's adept in code generation and completion, trained on an expansive dataset of 2 trillion tokens, primarily consisting of various programming languages and supplemented with English and Chinese natural language data. This extensive training allows the model to excel in tasks like repository-level code completion and project-level tasks, supported by a 16K context window. Instruction-tuned and quantized versions are available, balancing model size with performance, though quantization can affect quality. Despite its strengths, the model's focus is mostly on computer science, and it may offer technically correct yet impractical solutions, influenced by hardware capabilities.

Capabilities

Providers(1)