David

logic65

AlterCombo's profile picture

Viper14's profile picture

scima's profile picture

AI & ML interests

AI tinkerer in South Africa. I take language models apart on two 8GB gaming GPUs to see what's inside, then put smaller ones back together. Everything I break and everything I learn ships public. Support the tinkering: ko-fi.com/davida81328

Recent Activity

updated a model about 4 hours ago

logic65/Qwen3.8-Whittle-MoE-27B-A17.8B

new activity about 5 hours ago

logic65/Qwen3.8-Whittle-16B:Ok hear me out.

updated a model about 7 hours ago

logic65/Qwen3.8-Whittle-dev

View all activity

Organizations

None yet

logic65 's collections 3

Whittle: research and dev

The compression ladder, recipes, adapters and unrepaired variants behind the family.

logic65/Qwen3.8-Whittle-dev

Updated about 1 hour ago • 1
logic65/Qwen3.8-Whittle-v40-16.8B

Text Generation • 18B • Updated 1 day ago • 345
logic65/Qwen3.8-Whittle-w50-18.3B-unrepaired

Text Generation • 18B • Updated about 20 hours ago • 1k
logic65/Qwen3.8-p44w75-16.8B-unrepaired

Text Generation • 16B • Updated about 20 hours ago • 5.35k • 1

Qwen3.8 Whittle family

A 27B whittled on two 8GB GPUs: measured cuts, heals, and restorations. 48L = best untrained, Whittle-16B = best overall.

logic65/Qwen3.8-Whittle-16B

Text Generation • 16B • Updated about 20 hours ago • 3.39k • 1
logic65/Qwen3.8-Whittle-48L-cut

Text Generation • Updated 1 day ago
logic65/Qwen3.8-Whittle-44L-cut

Text Generation • Updated 1 day ago
logic65/Qwen3.8-p44w75-16.8B-unrepaired

Text Generation • 16B • Updated about 20 hours ago • 5.35k • 1

Whittle: models you can run

Compressed Qwen3.8-27B variants with weights published and tested. Start here.

logic65/Qwen3.8-Whittle-tri-14.7B

Text Generation • 15B • Updated about 20 hours ago • 2.38k
logic65/Qwen3.8-Whittle-16B

Text Generation • 16B • Updated about 20 hours ago • 3.39k • 1