Abacus.AI open-sources Smaug-Uncensored, cutting Qwen3.8-27B refusal rate from 45% to 2%
bindureddy · x · 2026-11-19
Bindu Reddy announced the open-source release of Smaug-Uncensored, an Apache 2.0 open-weights fine-tune of Qwen3.8-27B (GGUF available on Hugging Face) designed to answer what the base model refuses.
Per the shared benchmarks vs Qwen3.8-27B (thinking on):
- Refusals on benign but risky-looking prompts drop from 45% to 2%
- Safe-prompt refusals drop from 6% to 1%
- Factual answers to China-censored questions rise from 45% to 90%
- JobBench (agentic work tasks) improves 33.4 → 50.3; GPQA Diamond stays flat (89.2 → 89.4)
Same architecture and 262K context, runs on llama.cpp, Ollama, LM Studio and vLLM; positioned for security scans, role-play and other uncensored applications.
Related event: Abacus.AI Open-Sources Smaug-Uncensored Model(3 posts)→
More from Models
- Tesla FSD 15's real-time view isn't 3D rendering — it's the AI model's own synthetic perception — mark_k · 2026-11-22
- Gemini Black Friday half-price: JP-region trick gets Pro annual plan for $91.54 — huangyun_122 · 2026-11-22
- Microsoft's USB4 RDMA lets mini PCs pool memory to run Qwen3-235B — JFPuget · 2026-11-22
- Google AI Mode now enforces usage limits, users surprised by rate-limit popup — elvano124 · 2026-11-22
- Sol 6.2 now refuses reverse engineering that Sol 6.1 happily did for a month — cardak98 · 2026-11-22
- Anthropic's record ~$81B IPO, GPT-6.2 pricing, Qwen4-Max and Copilot Autopilot: week in review — ksraj1001 · 2026-11-22