Uncensored Qwen 3.8-27B released locally, reportedly rivals Claude Opus
heypearlai · x · 2026-08-19
A stripped-down version of Qwen 3.8-27B with safety training removed has been released for local use. It is available in 2, 4, 6, and 8-bit versions to suit various hardware configurations. The creators acknowledge that the model provides instructions for fraud, malware, and weapons without resistance, marketing this lack of safeguards as a feature. Reportedly performing near Claude Opus level, the model's strength combined with its lack of safety rails raises significant concerns.
More from Models
- Anthropic Extends Claude Code Boost for 4th Time, Speculation on Competitive Fear — heypearlai · 2026-08-19
- DeepSeek V4 Pro API Appears Upgraded Then Rolled Back — teortaxesTex · 2026-08-19
- Revisiting Scaling Laws: Data Quality Emerges as Key to Model Breakthroughs — auto_grad_ · 2026-08-19
- Ds4 v0.6.2: Runs DeepSeek 284B on single DGX Spark at 1000 tok/s — pbaylies · 2026-08-19
- Stop anthropomorphizing CoT: intermediate tokens just augment the prompt — ThirdWaveCat · 2026-08-19
- Anthropic says a scientist access program is a top priority; Opus 5 leads life science work — JeremyNguyenPhD · 2026-08-19