Sulat.com
AI models
NanoGPT logo

Model details

Qwen 3.8 27B Obliterated

Qwen 3.8 27B Obliterated is a community-distributed variant in the Qwen family designed to lower refusal behavior while preserving the underlying model's chat and reasoning strengths. Third-party listings describe it as a refusal-reduced Qwen 3.8 27B build intended for long-context chat and coding, with the ability to emit or hide thinking traces wrapped in think-style blocks and to support tight decoding controls. Its bf16 precision option and reasoning tag indicate it is well suited for structured, step-by-step problem solving rather than casual one-shot answers. The model was open-sourced on Hugging Face on August 20, 2026 by jailbreak researcher Pliny the Liberator, and it was the second independent uncensored Qwen3.8-27B build to surface, following OrcaRouter's earlier release a couple of days prior.

The Obliterated variant went viral in late August 2026 because it answered sensitive prompts directly, including a methamphetamine-synthesis explanation that ran successfully on an ordinary Mac without any jailbreak engineering, and the story gathered over 2,500 social posts within a day. That viral moment reflects how low the friction to local use has become, not a recommendation for misuse. Practically, the model is a good fit for red-teaming, safety evaluations, refusal-pattern research, and long-context coding or reasoning tasks where developers want maximum compliance with carefully scoped prompts. Teams looking for a Qwen-family chat model with open weights, optional visible reasoning, and relaxed refusal behavior will find it especially useful for adversarial testing and policy experimentation.

NanoGPTqwen/qwen3.8-27b-obliteratedqwen

Quick Info

Powered by
Provider
NanoGPT
Model key
qwen/qwen3.8-27b-obliterated
Release date
Jul 29, 2026
Last updated
Aug 24, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost
$0.25
Output token cost
$1.50

Limits

Input tokens
262,144 tokens
Output tokens
32,768 tokens
Context window
262,144 tokens

Latest news about Qwen 3.8 27B Obliterated

Videos about Qwen 3.8 27B Obliterated

Recent tweets and retweets from NanoGPT

More models around Qwen 3.8 27B Obliterated