deepseek-coder-6.7b-instruct-zse-int4

Pre-converted ZSE model for ultra-fast inference.

Source Model

Original: deepseek-ai/deepseek-coder-6.7b-instruct
Quantization: INT4
File Size: 3.61 GB
Format: ZSE binary (.zse)

Usage

pip install zllm-zse

# Download and serve
zse pull deepseek-coder-6.7b-instruct
zse serve deepseek-coder-6.7b-instruct

# Or direct
zse serve deepseek-coder-6.7b-instruct-zse-int4.zse

Benefits

5x faster cold start compared to HuggingFace loading
10-14% less VRAM with ZSE custom INT4 kernels
Single file — tokenizer and config embedded
No internet required after download

Benchmarks

See ZSE Documentation for full benchmarks.

Converted with ZSE v1.4.0

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for zse-zllm/deepseek-coder-6.7b-instruct-zse-int4

Base model

deepseek-ai/deepseek-coder-6.7b-instruct

Finetuned

(60)

this model