Qwen3.8-27B Uncensored Runs on Macs

// 46d agoMODEL RELEASE

Qwen3.8-27B Uncensored Runs on Macs

Community-built uncensored variants of Qwen3.8-27B are now available in MLX-compatible quantizations for Apple Silicon, enabling private local inference without CUDA or cloud APIs. The model retains its reasoning, vision, video, and agentic capabilities.

// ANALYSIS

This is a strong win for Mac-based local AI, though community quantizations still need independent quality and safety testing.

–3.8-bit and 4.95-bit MLX quants target practical Apple Silicon memory limits
–Local execution keeps prompts, files, and tool workflows off third-party servers
–Retaining multimodal and reasoning capabilities makes it more useful than a stripped-down text-only uncensored model
–Uncensored fine-tuning removes refusal behavior but can also remove important safety guardrails
–Early community builds may vary significantly in speed, accuracy, compatibility, and quantization quality

// TAGS

qwen3.8-27b-uncensoredllmopen-weightsquantizationinferenceedge-ailocal-firstvision

DISCOVERED

46d ago

2026-08-19

PUBLISHED

46d ago

2026-08-19

RELEVANCE

9/ 10

AUTHOR

WorldofAI

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

RESEARCH24m ago

Finetuning with Sampling Makes SFT Rival RL

Harvard researchers introduce an MCMC-based projection-sampling method that reshapes expert off-policy traces toward a model’s own distribution before supervised fine-tuning. Across chemistry, math, and medical QA tasks, the approach reportedly improves generalization and reduces forgetting versus strong RL and self-distillation baselines.

RESEARCH24m ago

Sharpening Tax Exposes RL’s Coverage Trade-Off

This research finds that RL post-training can improve pass@1 while narrowing an LLM’s solution coverage under repeated sampling. It introduces Sharpening Tax to measure that loss and PTGS, which adapts training temperature to prompt difficulty.

OPEN SOURCE56m ago

Crawlers Debuts Open Crawler Sandbox

Crawlers is an open-source, MIT-licensed browser sandbox for configuring autonomous crawler swarms, running them through a deterministic synthetic web, and inspecting or forking each simulation locally.