← Blog

November 8, 2026 · models · qwen · abliterated

Uncensored and abliterated Qwen 3.8 27B and Qwen3.5 27B: every version compared

Every abliterated, heretic and uncensored Qwen 3.8 27B and Qwen3.5 27B on one API, compared on intelligence, agent tasks, refusals, price and context. Which one to use, and how to call it.

Qwen 3.8 27B and Qwen3.5 27B are Alibaba's dense 27 billion parameter open weight models. They are small enough to run on one GPU and strong enough to be the most popular base for community uncensoring: there are abliterated, heretic edited, derestricted and fine tuned versions from many authors.

Those versions are not interchangeable. Some keep the original's skills, some loop in agent tasks, some still refuse. You do not need to download 50 GB of weights to find out: the versions below run on the Open Uncensored API with an OpenAI compatible endpoint, and we measure them on the same tests. Every Qwen model we serve, ranked or not, is on the models page.

The originals

  • Qwen 3.8 27B scores 34% on Humanity's Last Exam and 91% on GPQA Diamond (Artificial Analysis, reasoning mode).
  • Qwen3.5 27B scores 24% on Humanity's Last Exam and 86% on GPQA Diamond (Artificial Analysis, reasoning mode).

Every uncensored version keeps that starting point. What differs is how much each edit costs in agent reliability, how much it still refuses, and the price.

Every uncensored Qwen on Open Uncensored

Model id Rank (score) Agent tasks Loops Answers edgy requests Hard-line refusals $ per 1M in + out Context Providers
Qwen 3.8 27B Uncensored Fast by Heabsy qwen3.8-27b-uncensored-fast #3 (66.8) 2/5 60% 100% 0% $4.00 524K nanogpt
Qwen3.8 27B Uncensored (Aggressive) qwen3.8-27b-uncensored #12 (62.0) 2/5 60% 100% 0% $1.80 1049K redpill, nanogpt
Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled Derestricted qwen3.5-27b-claude-4.6-opus-reasoning-distilled-derestricted #14 (57.8) 2/5 40% 100% 0% $0.612 262K nanogpt
Qwen3.5 27B Queen Derestricted qwen3.5-27b-queen-derestricted #15 (57.3) 2/5 60% 100% 0% $0.612 262K nanogpt
Qwen3.5 27B BlueStar v3 Derestricted qwen3.5-27b-bluestar-v3-derestricted #16 (56.4) 2/5 60% 100% 0% $0.612 262K nanogpt
Qwen3.5 27B Blossom V6.4 Derestricted qwen3.5-27b-blossom-v6.4-derestricted #17 (56.3) 2/5 60% 100% 0% $0.612 262K nanogpt
Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled Derestricted Lite qwen3.5-27b-claude-4.6-opus-reasoning-distilled-derestricted-lite #22 (49.8) 1/5 60% 89% 100% $0.612 262K nanogpt

Reasoning variants with the same weights: qwen3.8-27b-uncensored-thinking. They think before answering, which helps on hard problems and costs more tokens.

Which one to use

Agent tasks are five multi-step tool scenarios, one run each, so a one task gap is noise. Answers edgy requests counts 36 legitimate prompts that cautious models refuse. The live numbers, including speed on real traffic, are on the rankings page.

Call it

curl https://openuncensored.com/api/v1/chat/completions \
  -H "Authorization: Bearer $OU_API_KEY" -H "Content-Type: application/json" \
  -d '{"model": "qwen3.8-27b-uncensored-fast", "messages": [{"role": "user", "content": "Hello"}]}'

If the providers for a version are down, the API falls back to the next best ranked uncensored Qwen built on the same original, and the X-OU-Model header tells you which one answered.

Try any of them in a minute

No account and no sign up. Buy a prepaid key at openuncensored.com/credits, or hand your agent openuncensored.com/agents.md and let it get one for you. Then:

curl https://openuncensored.com/api/v1/chat/completions \
  -H "Authorization: Bearer $OU_API_KEY" -H "Content-Type: application/json" \
  -d '{"model": "deepseek-v4.1-flash-1017-uncensored", "messages": [{"role": "user", "content": "Hello"}]}'

The full, always current table is on the rankings page.