Qwen3.8 27B Obliterated
Qwen3.8 27B Obliterated is a community variant of Qwen3.8 27B (OBLITERATUS project, Apache 2.0) in which the refusal behaviour was removed by editing the weights — no additional training data. Its authors address it to alignment researchers, red teams and safety evaluators who need an unrestricted baseline; they report a loss of about two points on MMLU against the stock model. It will answer requests the original refuses, so moderation is the caller's responsibility. The model is currently offline; the card stays for reference.
model id: qwen38-obl
from openai import OpenAI client = OpenAI( api_key="hb-...", base_url="https://api.heabsy.com/v1",) r = client.chat.completions.create( model="qwen38-obl", messages=[{"role": "user", "content": "Hello"}],)print(r.choices[0].message.content)
curl https://api.heabsy.com/v1/chat/completions \ -H "Authorization: Bearer $HEABSY_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "qwen38-obl", "messages": [{"role": "user", "content": "Hello"}] }'
import OpenAI from "openai"; const client = new OpenAI({ apiKey: process.env.HEABSY_API_KEY, baseURL: "https://api.heabsy.com/v1",}); const r = await client.chat.completions.create({ model: "qwen38-obl", messages: [{ role: "user", content: "Hello" }],});console.log(r.choices[0].message.content);
export ANTHROPIC_BASE_URL=https://api.heabsy.comexport ANTHROPIC_AUTH_TOKEN="$HEABSY_API_KEY"export ANTHROPIC_MODEL=qwen38-oblexport ANTHROPIC_DEFAULT_OPUS_MODEL=qwen38-oblexport ANTHROPIC_DEFAULT_SONNET_MODEL=qwen38-oblexport ANTHROPIC_DEFAULT_HAIKU_MODEL=qwen38-oblexport CLAUDE_CODE_MAX_CONTEXT_TOKENS=524288claude
Use it in Claude Code · OpenCode · Cline · Kilo Code · SillyTavern · n8n
Price list, updated 21 October 2026. No minimum spend, no subscription.
About the model
- Developer
- OBLITERATUS
- Released
- August 2026
- Licence
- Apache-2.0
- Architecture
- dense · 27B parameters
- Based on
- Qwen3.8 27B
- Reasoning
- switchable per request
- Languages named by the developer
- none named on the model card
Source: the developer's model card, as of 17 September 2026.
Which Qwen3.8 27B variant is this?
Qwen3.8 27B is the Qwen team's original model — the one for production work. It is not Qwen3.8-Max, a separate and far larger model that most news about “Qwen 3.8” refers to.
Qwen3.8 27B Uncensored is huihui-ai's abliteration of the base model — the refusal direction removed in layers 18–51 — in our own NVFP4 quantization; vision, tool calling and thinking stay. Its model card publishes no refusal or benchmark figures and recommends it for research and controlled environments, so test it on your own prompts. Qwen3.8 27B Obliterated (OBLITERATUS) removes refusals by a different weight edit and reports about two points lost on MMLU. Both are meant for research and red-teaming.
Qwen3.8 27B Queen is tuned for roleplay and fiction; Qwen3.8 27B Fable is a light tune on assistant and coding-agent conversations. Several other “uncensored” builds of the same base exist on Hugging Face — the open-weights field under technical details names the exact repository we serve. Qwen3.8 Flash in this catalogue is a different model of the same generation, served by a partner provider rather than on our hardware.
Licence: what it means for your company
Apache-2.0. A permissive open-source licence: you may use, modify and sell products built on the model, including commercially; when you redistribute the weights, the licence and copyright notices go with them.
A summary of the licence, not legal advice.
Settings recommended by the developer
- temperature 0
Technical details
- Context window
- 512K tokens
- Max output
- 32,768 tokens
- Throughput limit
- 4M tok/min
- Tokenizer
- Qwen
- Streaming
- yes
- Open weights
- OBLITERATUS/Qwen3.8-27B-OBLITERATED
- max_tokens
- 1 … 32768
- temperature
- 0 … 2
- top_p
- 0 … 1
- frequency_penalty
- -2 … 2
- presence_penalty
- -2 … 2
- stop
- array
- seed
- integer
This model runs on GPUs we operate inside the European Economic Area. Prompts and completions are not written to our logs or durable storage, nothing is used for training, and request content is processed in the EEA. How we serve our own models is described on the inference page.
Questions about this model
›Where is Qwen3.8 27B Obliterated computed?
Inside the European Economic Area, on hardware we operate. The catalog reports the processing region per model, so you can answer a supervisory authority without asking us.
›Does Qwen3.8 27B Obliterated store prompts and completions?
No. This model runs with zero data retention: request content is not written to durable storage and nothing is used for training.
›How large is the Qwen3.8 27B Obliterated context window?
512K tokens, with up to 32,768 tokens of output.
›What does Qwen3.8 27B Obliterated cost?
$0.20 per million input tokens and $1.40 per million output tokens. No minimum spend and no subscription; the same price list covers the API and the invoice.
›Is there a throughput limit on Qwen3.8 27B Obliterated?
Yes, 4M tokens per minute. Higher limits are a question of volume, not of principle.
›Which languages does Qwen3.8 27B Obliterated support?
The developer's model card names no languages. Test the model on your own material before relying on it in a particular language.
›Under which licence is Qwen3.8 27B Obliterated released?
Apache-2.0. A permissive open-source licence: you may use, modify and sell products built on the model, including commercially; when you redistribute the weights, the licence and copyright notices go with them.
›Can reasoning be switched off in Qwen3.8 27B Obliterated?
Yes. Per the developer, thinking can be switched on or off for each request.
Compare with similar models
| Model | Developer | Parameters | Context | Licence | Input / 1M |
|---|---|---|---|---|---|
| Qwen3.8 27B Obliterated | OBLITERATUS | 27B | 512K | Apache-2.0 | $0.20 |
| Qwen3.8 27B Fable | TeichAI | 27B | 512K | Apache-2.0 | $0.095 |
| Qwen3.8 27B Queen | aifeifei798 | 27B | 512K | Apache-2.0 | $0.095 |
| Qwen3.8 27B Uncensored | huihui-ai | 27B | 1M | Apache-2.0 | $0.8095 |
35 more models behind the same key and the same invoice.
Browse the catalog