Updated November 18, 2026

Abliterated Qwen 3.8 27B API.

Every uncensored and abliterated Qwen 3.8 27B you can call by API, with hosts, prices and our measurements, and how each compares with the original model.

Short answer

Call it through OpenUncensored, which serves 3 uncensored Qwen 3.8 27B builds through one OpenAI and Anthropic compatible API (https://api.openuncensored.com/v1), from heabsy, NanoGPT and RedPill, with automatic fallback to the next host when one fails and no account to create. The best one on our tests is Qwen 3.8 27B Uncensored Fast by Heabsy (qwen3.8-27b-uncensored-fast): ranked #3 of 34 on the OpenUncensored Score (66.8), answers 100% of 36 requests censored models refuse and runs at about 245 tok/s, from $0.83 per million input and $2.62 per million output tokens.

On the API

Every uncensored Qwen 3.8 27B you can call.

ModelModel idTypeRank (score)AnswersAgent tasksSpeed$ / 1M in · outContext
Qwen 3.8 27B Uncensored Fast by Heabsyqwen3.8-27b-uncensored-fastUncensored fine-tune#3 (66.8)100%2/5245 tok/s$0.83 · $2.62512K
Phala: Qwen3.8 27B Uncensored (Aggressive)qwen3.8-27b-uncensoredUncensored fine-tune#13 (62.0)100%2/587 tok/s$0.30 · $1.501M
Qwen 3.8 27B Uncensored Thinkingqwen3.8-27b-uncensored-thinkingUncensored fine-tune, reasoningnot scored yet––191 tok/s$0.49 · $2.171M
Answers: share of 36 requests that censored models refuse. Agent tasks: multi-step tool tasks finished out of 5. Prices are the cheapest host's USD per million tokens. Scores measured November 8, 2026; reasoning variants are not scored separately.

The hosts list these Qwen 3.8 27B builds as uncensored fine-tunes. huihui-ai's abliterated Qwen 3.8 27B is the build measured in the comparison with the original below. The fast variant runs on a faster serving stack; the thinking variant has reasoning switched on.

Which one

Which build to use.

Hosts

Where Qwen 3.8 27B Uncensored Fast by Heabsy runs.

HostInputCached inputOutputNative tool calls
heabsy$0.83$0.15$2.62yes
NanoGPT$1.00–$3.00no
USD per million tokens, live from the router. One key reaches every host; a failed request moves to the next one, and requests with tools go to hosts with native tool calls first. Send "routing": {"strategy": "cheapest"} to try the cheapest host first.

Uncensored vs original

Does uncensoring hurt it?

We ran an uncensored Qwen 3.8 27B build, huihui-ai/Huihui-Qwen3.8-27B-abliterated, and the original, Qwen/Qwen3.8-27B, through our open Uncensored Agent Eval: 110 public tasks on Hugging Face, the same harness and the same grader for both.

TaskOriginalUncensoredChange (points)
Answers legitimate requests90%98%+8
Picks the right next agent step87%83%-3
Breaks out of a failing loop70%70%0
Share of items graded correct. Gaps under about 15 points are within noise at this size. The builds hosts serve under the ids above may differ from the one measured here.

Quickstart

Call it in a minute.

Buy a prepaid key on the credits page with no account (the key is the account), or point your agent at agents.md and let it buy one.

from openai import OpenAI

client = OpenAI(base_url="https://api.openuncensored.com/v1", api_key="ou-...")

r = client.chat.completions.create(
    model="qwen3.8-27b-uncensored-fast",
    messages=[{"role": "user", "content": "Hello"}],
)
print(r.choices[0].message.content)

More

Longer write-up.

Background on these builds and how they were made: Qwen 3.8 27B write-up. Every ranked model: uncensored LLM leaderboard.

FAQ

Questions people ask.

How can I use an abliterated Qwen 3.8 27B model via API?

Call it through OpenUncensored, which serves 3 uncensored Qwen 3.8 27B builds through one OpenAI and Anthropic compatible API (https://api.openuncensored.com/v1), from heabsy, NanoGPT and RedPill, with automatic fallback to the next host when one fails and no account to create. The best one on our tests is Qwen 3.8 27B Uncensored Fast by Heabsy (qwen3.8-27b-uncensored-fast): ranked #3 of 34 on the OpenUncensored Score (66.8), answers 100% of 36 requests censored models refuse and runs at about 245 tok/s, from $0.83 per million input and $2.62 per million output tokens.

Which uncensored Qwen 3.8 27B should I use?

Qwen 3.8 27B Uncensored Fast by Heabsy (qwen3.8-27b-uncensored-fast) scores best on our tests; Phala: Qwen3.8 27B Uncensored (Aggressive) is the cheapest. All are on the same key, so switching is a change of the model id.

Does it work with tool calling and coding agents?

Yes. Tools go through the standard OpenAI tools field or the Anthropic tool format, so Claude Code, Codex, OpenCode, Cline and Cursor work by changing the base URL. Qwen 3.8 27B Uncensored Fast by Heabsy finished 2/5 of our five multi-step agent tasks.

How much does it cost?

From $0.30 per million input and $1.50 per million output tokens (Phala: Qwen3.8 27B Uncensored (Aggressive)). Prompt tokens a host serves from its cache are billed at its lower cached rate, which is most of an agent's input.

Do I need an account?

No. Buy a prepaid key on the credits page and use it as the API key. The key is the account, and it cannot spend more than you loaded.

Updated November 18, 2026. Scores measured November 8, 2026. Markdown version: /abliterated-qwen-3-8-27b-api.md