---
license: apache-2.0
language:
- en
- zh
pipeline_tag: text-generation
tags:
- apple-silicon
- metal
- splash
- mtp
- speculative-decoding
- 8bit
---

# Qwen3.8-27B-Splash-Q8 (Compressed 8-bit Baseline)

This repository contains the baseline 8-bit compressed weights for **Qwen3.8-27B** for the Splash inference engine on Apple Silicon.

## Performance
* **Decode Speed**: 36.5 tok/s average (peaking at 52.7 tok/s).
* **Speedup**: 3.69x over standard autoregressive decoding.
* **Reasoning Accuracy**: 44.4% across GPQA Diamond and AIME 2025.