Shadowgraph
TrendingBuzzAbout
Sign in

nitinpanj/Qwen3.8-27B-Splash-Q8

Model

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

1

8 commits

1 linked in READMEs

updated Sep 20, 2026

See the code

README

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.
8bit
apple-silicon
metal
speculative-decoding
splash
text-generation

Similar repos

nitinpanj/Qwen3.8-27B-Splash-Mixed

Model

Qwen3.8-27B-Splash-Mixed (Selective 8-bit Sensitive Layers)

1

nitinpanj/Qwen3.8-27B-Splash-HQ

Model

Qwen3.8-27B-Splash-HQ (Native 8-bit)

7

incoai/Qwen3.8-27B-Splash

Model

Qwen3.8-27B-Splash

132

incoai/Qwen3.6-35B-A3B-Splash

Model

Qwen3.6-35B-A3B-Splash

31

npanj/splash-plus

Empirical benchmarks and quality evaluations of Qwen3.8-27B with Multi-Token Prediction (MTP)…

13

mlboydaisuke/Qwen3.8-27B-CoreAI

Model

Core AI is Apple's on-device ML runtime in iOS 27 / macOS 27 and the successor to Core ML: PyTorch…

1

mlboydaisuke/qwen3.5-0.8B-CoreAI

Model

Core AI is Apple's on-device ML runtime in iOS 27 / macOS 27 and the successor to Core ML: PyTorch…

0

Qwen/Qwen3.8-27B-FP8

Model

Qwen3.8-27B-FP8

880

Linked in READMEs

npanj/splash-plus

Empirical benchmarks and quality evaluations of Qwen3.8-27B with Multi-Token Prediction (MTP)…

13

README badge

nitinpanj/Qwen3.8-27B-Splash-Q8

Model

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

1

8 commits

1 linked in READMEs

updated Sep 20, 2026

See the code

README

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.
8bit
apple-silicon
metal
speculative-decoding
splash
text-generation

Similar repos

nitinpanj/Qwen3.8-27B-Splash-Mixed

Model

Qwen3.8-27B-Splash-Mixed (Selective 8-bit Sensitive Layers)

1

nitinpanj/Qwen3.8-27B-Splash-HQ

Model

Qwen3.8-27B-Splash-HQ (Native 8-bit)

7

incoai/Qwen3.8-27B-Splash

Model

Qwen3.8-27B-Splash

132

incoai/Qwen3.6-35B-A3B-Splash

Model

Qwen3.6-35B-A3B-Splash

31

npanj/splash-plus

Empirical benchmarks and quality evaluations of Qwen3.8-27B with Multi-Token Prediction (MTP)…

13

mlboydaisuke/Qwen3.8-27B-CoreAI

Model

Core AI is Apple's on-device ML runtime in iOS 27 / macOS 27 and the successor to Core ML: PyTorch…

1

mlboydaisuke/qwen3.5-0.8B-CoreAI

Model

Core AI is Apple's on-device ML runtime in iOS 27 / macOS 27 and the successor to Core ML: PyTorch…

0

Qwen/Qwen3.8-27B-FP8

Model

Qwen3.8-27B-FP8

880

Linked in READMEs

npanj/splash-plus

Empirical benchmarks and quality evaluations of Qwen3.8-27B with Multi-Token Prediction (MTP)…

13

README badge