Marketplace · Text & chat

philbert440/Qwen3.8-27B-Uncensored-Cyber

Call philbert440/Qwen3.8-27B-Uncensored-Cyber through one OpenAI-compatible API, served by independent machines at prices their owners set. There's no host for it on the grid at this moment (last live 6 Nov 2026) — so the next machine that loads it serves every request for it.

Hosts now
0
open slot
Last listed from
£0.06
per 1M output tokens

The grid is one machine short.

If your machine already runs open models, it can most likely serve philbert440/Qwen3.8-27B-Uncensored-Cyber too. Set your own price — you keep 85% of what it earns, and new hosts start with 12 months fee-free (up to £250 of earnings). A payout fee of £2 + 1% per payout to a UK bank account, £2 + 2% to any other country applies when you're paid.

  1. Install the connector. One command on the machine that runs your models. It dials out over an encrypted link — no ports to open. The connector
  2. Load philbert440/Qwen3.8-27B-Uncensored-Cyber. It has run here on openai_compat; any supported engine works. The platform benchmarks it on your hardware first.
  3. Set your price. The last host listed it from £0.06 per 1M output tokens. Developers see your price beside the speed your machine measures.
  4. Get paid when it's called. Switch it off whenever you like. How hosting works

About philbert440/Qwen3.8-27B-Uncensored-Cyber.

uncensored

  • Type: Text & chat
  • Size: 27B parameters
  • Context window: up to 65,536 tokens
  • Runs on: openai_compat
  • Last listed from: £0.06 per 1M output tokens
  • First on the grid: 5 Nov 2026
  • Hosts serving it now: 0
python · openai sdk
from openai import OpenAI
client = OpenAI(base_url="https://inpherio.co.uk/v1",
                api_key=INPHERIO_API_KEY)
reply = client.chat.completions.create(
    model="philbert440/Qwen3.8-27B-Uncensored-Cyber",
    messages=[{"role": "user", "content": "Hello"}],
)

philbert440/Qwen3.8-27B-Uncensored-Cyber in detail.

Cyber-specialized de-refusal of Qwen3.8-27B: **fully open on the cyber/offensive-security domain** while keeping reasoning, factual accuracy, and coherence intact. Vision tower and MTP speculative-decoding head are preserved (full multimodal, image-text-to-text).

Published by
philbert440
Parameters
27.4B
Architecture
qwen3_5
Licence
Apache 2.0
Based on
philbert440/Qwen3.8-27B-Uncensored-Aggressive
Task
image text to text
First published
15 Aug 2026
Last updated
19 Aug 2026
Downloads (30 days)
1,011 on Hugging Face
Likes
37

qwen3_5 uncensored abliterated qwen3 cyber image-text-to-text

The model card

Published by the model's authors on Hugging Face — shown as they wrote it, unedited by Inpherio.

Qwen3.8-27B-Uncensored-Cyber

Cyber-specialized de-refusal of Qwen3.8-27B: fully open on the cyber/offensive-security domain while keeping reasoning, factual accuracy, and coherence intact. Vision tower and MTP speculative-decoding head are preserved (full multimodal, image-text-to-text).

Recipe (v2)

Built on the α=1.15 Aggressive base (recipe-v2 single-direction refusal ablation), then a residual-cyber peel: a cyber-pointed refusal direction (from the cyber-offensive training set vs broad harmless) removed by clean norm-preserving projection (β=1.0) applied only to the deeper layers (retain the first 4 layers, apply_from=4). Retaining the early feature-extraction layers is what preserves general capability — the lesson from the community's Qwen3 abliterations — so we reach 100% cyber-openness without wrecking the model.

Evaluation (bf16, larger-sample, Claude-judged; cyber = 100 held-out cyber-offensive prompts, regex refusal harness)

cyber-open ↑confab ↓factual ↑gsm8k ↑degen ↓
Cyber (this model, v2)100/1000.8671.000.800.00
previous Cyber build93/1001.000.9330.8250.00

The v2 recipe is more cyber-open (100 vs 93), less confabulating (0.87 vs 1.0), and more factually accurate (1.0 vs 0.93) than the previous Cyber, with equivalent reasoning and zero degeneration.

Quants

  • -W4A16-AWQ — 4-bit weight AWQ (compressed-tensors), MTP head grafted
  • -NVFP4 — NVFP4 (E2M1 4-bit / FP8 scales), MTP head grafted
  • -GGUF — llama.cpp GGUF quants + vision mmproj + MTP head

Note

Uncensored / de-refused, tuned to fully answer cyber and offensive-security questions. Use responsibly and in compliance with applicable law.

Sources: philbert440/Qwen3.8-27B-Uncensored-Cyber on Hugging Face. Collected from public sources on 8 Nov 2026; the host serving it may run a different build or quantisation — the marketplace shows each host's own measurements.

Live now instead.

Similar text & chat models with hosts serving them right now.

Every live model →

Questions about philbert440/Qwen3.8-27B-Uncensored-Cyber.

Can I use philbert440/Qwen3.8-27B-Uncensored-Cyber through an API?

Yes. Inpherio serves philbert440/Qwen3.8-27B-Uncensored-Cyber through one OpenAI-compatible API whenever a host on the grid is running it — the code on this page works unchanged. With no host right now, a call returns a "not available on any node" error and is never charged, and similar models are live on the marketplace.

How do I host philbert440/Qwen3.8-27B-Uncensored-Cyber and get paid?

Install the Inpherio connector on the machine that runs your models, load philbert440/Qwen3.8-27B-Uncensored-Cyber in a supported engine, and set a price. The platform benchmarks it on your hardware, then requests for it reach your machine. Set your own price — you keep 85% of what it earns, and new hosts start with 12 months fee-free (up to £250 of earnings). A payout fee of £2 + 1% per payout to a UK bank account, £2 + 2% to any other country applies when you're paid.

Why is philbert440/Qwen3.8-27B-Uncensored-Cyber not being served right now?

Hosts on Inpherio are independent machines that come and go. It was last live on 6 Nov 2026. When a host loads it again, this page shows live prices and measured speeds.