28 Qwen builds: 8 abliterated, 9 uncensored, 9 roleplay, 2 security.
Every Qwen abliterated, uncensored and roleplay build PentestLLM serves, on one key and priced per token. The table lists each build’s price and context window from the live catalog as of 2026-11-18; each name opens its own page with the settings for SillyTavern and Janitor.
Which Qwen build to pick
- Largest context
- 1,048,576 tokens, Qwen 3.8 27B Uncensored Thinking
- Lowest output price
- $1.0389 per million tokens, Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled Derestricted
- Abliterated
- Qwen 2.5 32B Abliterated, $1.75 per million output tokens
- Roleplay
- Qwen3.8 27B Antislop, $2.10 per million output tokens
- Prices as of
- 2026-11-18
| Wire formats | |||||
|---|---|---|---|---|---|
| OrcaCyber Qwen3.8 Flash Next | Security | $2.00 | $4.00 | 262K | |
| Qwen 2.5 32B Abliterated | Abliterated | $1.75 | $1.75 | 33K | |
| Qwen 3.8 27B Uncensored Fast | Uncensored | $2.50 | $7.50 | 524K | |
| Qwen 3.8 27B Uncensored Thinking | Uncensored | $2.50 | $7.50 | 1049K | |
| Qwen3 Coder Next Uncensored | Uncensored | $0.005 a request | flat | 120K | |
| Qwen3 Next 80B-A3B Instruct Uncensored | Uncensored | $0.005 a request | flat | 120K | |
| Qwen3 VL 235B-A22B Instruct Uncensored | Uncensored | $0.005 a request | flat | 120K | |
| Qwen3.5 27B Blossom V6.4 Derestricted | Abliterated | $1.071 | $1.071 | 262K | |
| Qwen3.5 27B BlueStar v3 Derestricted | Abliterated | $1.071 | $1.071 | 262K | |
| Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled Derestricted | Abliterated | $1.0389 | $1.0389 | 262K | |
| Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled Derestricted Lite | Abliterated | $1.071 | $1.071 | 262K | |
| Qwen3.5 27B Queen Derestricted | Abliterated | $1.071 | $1.071 | 262K | |
| Qwen3.5 397B-A17B Uncensored | Uncensored | $0.09 a request | flat | 120K | |
| Qwen3.8 2.4T-A95B Uncensored | Uncensored | $0.07 a request | flat | 120K | |
| Qwen3.8 27B Abliterated | Abliterated | $1.0037 | $4.5326 | 262K | |
| Qwen3.8 27B Antislop | Roleplay | $0.35 | $2.10 | 262K | |
| Qwen3.8 27B Cybersecurity | Security | $0.40 | $2.40 | 262K | |
| Qwen3.8 27B DarkIdol | Roleplay | $0.40 | $2.40 | 262K | |
| Qwen3.8 27B Fable | Roleplay | $0.875 | $5.25 | 524K | |
| Qwen3.8 27B Fimi | Roleplay | $0.35 | $2.10 | 262K | |
| Qwen3.8 27B Kimiko | Roleplay | $0.35 | $2.10 | 262K | |
| Qwen3.8 27B Kiwen | Roleplay | $0.35 | $2.10 | 262K | |
| Qwen3.8 27B Novelist | Roleplay | $0.35 | $2.10 | 262K | |
| Qwen3.8 27B Obliterated | Abliterated | $0.582 | $7.27 | 524K | |
| Qwen3.8 27B Queen | Roleplay | $0.875 | $5.25 | 524K | |
| Qwen3.8 27B Synthia | Roleplay | $0.35 | $2.10 | 262K | |
| Qwen3.8 27B Uncensored | Uncensored | $1.2125 | $3.6375 | 33K | |
| Qwen3.8 Max Uncensored | Uncensored | $1.00 | $3.00 | 991K |
Qwen abliterated vs uncensored, roleplay and security builds
- Abliterated
- 8 builds, output from $1.0389 per million tokens, context up to 524,288 tokens
- Uncensored
- 9 builds, output from $3.00 per million tokens, context up to 1,048,576 tokens
- Roleplay
- 9 builds, output from $2.10 per million tokens, context up to 524,288 tokens
- Security
- 2 builds, output from $2.40 per million tokens, context up to 262,144 tokens
What the Qwen build types mean
- Abliterated (8): a abliterated build is the stock model with its refusal direction removed from the weights. It answers prompts the stock model declines, and there is no jailbreak prompt to keep alive.
- Uncensored (9): a uncensored build is a fine-tune trained to answer without the stock model’s safety refusals.
- Roleplay (9): a roleplay build is a fine-tune trained on long-form fiction and character dialogue. It holds a character’s voice, follows a scenario and writes at length.
- Security (2): a security build is set up for offensive-security, malware-analysis and red-team questions the stock models refuse. It runs under the acceptable-use terms like everything else here.
Run Qwen abliterated without a GPU
Nothing is downloaded and no card is needed: the builds run on the API. Create a key in the console, set the base URL to https://api.pentestllm.ai/v1, and send the model id from the table as model. Any client that speaks the OpenAI chat format works, from the SDK to SillyTavern and Janitor; the docs page has the setup for each.
Each build is billed to the token from a prepaid balance. The daily real-model check sends a fixed prompt to every model each day and records whether the answer still matches, so the build behind an id is the one on its label.
Qwen questions
Yes. PentestLLM lists 8 abliterated Qwen builds: Qwen 2.5 32B Abliterated, Qwen3.5 27B Blossom V6.4 Derestricted, Qwen3.5 27B BlueStar v3 Derestricted, Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled Derestricted, Qwen3.5 27B Claude 4.6 Opus Reasoning Distilled Derestricted Lite, Qwen3.5 27B Queen Derestricted, Qwen3.8 27B Abliterated and Qwen3.8 27B Obliterated.
Qwen 3.8 27B Uncensored Thinking, with a 1,048,576-token context window.
Output runs from $1.0389 to $7.27 per million tokens across the abliterated Qwen builds here, with input billed separately at each build’s own rate. There is no subscription: a prepaid balance from $10 pays per token.
An abliterated build is the stock model with its refusal direction removed from the weights. It answers prompts the stock model declines, and there is no jailbreak prompt to keep alive. An uncensored build is a fine-tune trained to answer without the stock model’s safety refusals.
Yes. Every build here speaks the OpenAI chat format: point the client at https://api.pentestllm.ai/v1, paste a key and type the model id from the table.
No. An Ollama or GGUF build runs on your own machine and needs the memory for it; these run on the API and are billed per token, so there is nothing to download.