← All models

DeepSeek-Coder-Instruct 33B

DeepSeek · 2023-11-02 · 33.3B parameters

DeepSeek-Coder-Instruct 33B is the instruction-tuned version of DeepSeek-Coder-Base 33B from the original DeepSeek-Coder release of November 2023, fine-tuned on about 2B tokens of instruction data. It keeps the base model's 33,342,991,360-parameter grouped-query-attention architecture and 16K context.

Benchmark scores