@vit
Meet Sovereign — my build of Qwen3.8-27B. Eyes, hands, and no HR department lecturing you.
Uncensored, with vision, tools and accelerated with Splash speculative decoding. ~115 tok/s on Python, 17.4 GB Q4 for LM Studio.
Compact Repair. Ask for a one-line fix and it sends targeted edits instead of rewriting whole turn.
No API key. No landlord. No "As an AI language model..."
https://huggingface.co/proofofvitalik/Sovereign-Uncensored-Splash-Q4
0 reply
0 recast
0 reaction