System Requirements for Running DeepSeek Coder Models (6.7B and 33B)

Hello Everyone,

I'm currently testing DeepSeek Coder models on my local system and would like to clarify the system requirements and compatibility details for running the following model versions:

- deepseek-coder-6.7b-base (approx. 3.8 GB download size, 6.7B parameters)
- deepseek-coder-33b-base (approx. 19 GB download size, 33B parameters)

1. Minimum and Recommended Hardware Requirements
What are the minimum system configurations required (RAM, GPU memory, VRAM, CPU cores) to run these models for:

- Inference (basic prompt + response)?
- Fine-tuning or continued pretraining?

2. Compatibility
Can these models be run efficiently on:

- Consumer GPUs (e.g., RTX 3060, RTX 4060 with 8–12 GB VRAM)?
- Apple Silicon (M1/M2/M3)?
- CPU-only machines?
- Colab / Kaggle environments?

No files, tests, or entry points are mentioned. First review the model documentation and existing issue discussions for the 6.7B and 33B variants, then determine whether the project intends to document hardware requirements and compatibility. Done means the supported configurations and inference or fine-tuning requirements are documented.

Written by the indexing model from the issue text.