ThoxHeretic-27B

Public Community Redistribution (Not THOX-Trained)

27B · Q4_K_M GGUF + vision projector · Community redistribution — not THOX-trained · THOX evaluation: none published, no quality claim

Your AI. Your Data. Your Rules.™

Status: COMMUNITY REDISTRIBUTION — NOT THOX-TRAINED. This repository redistributes a community uncensored (abliterated) build: the Q4_K_M GGUF and its vision projector are in this repo (see Files). THOX did not train these weights, publishes no evaluation for them, and makes no quality claim. The model is not safety-tuned and is not part of the ThoxOS or MeshStack catalog.

Summary

Safety notice. This repo redistributes a community uncensored (abliterated) model. It is not trained by THOX, is not safety-tuned, and is not part of the ThoxOS or MeshStack catalog, and is not promoted for THOX devices. It has reduced refusal behaviour and must not be deployed to end users without your own safety layer. This repository is fully public and ungated: anyone can download these files, with no approval or account action required.

ThoxHeretic-27B is a multimodal GGUF redistribution (Q4_K_M, with a bundled vision projector file) made available for research and internal evaluation. The base version recorded in this card's metadata is not independently verified by THOX.

Founder's note — on creative freedom

THOX builds without artificial restriction. Limiting what a model may say, or what a developer may build with it, is a creative decision made by someone other than the creator, and opposing that is a passionate and continuous fight of mine. It shapes everything we publish and every tool we ship: open weights, honest documentation, and the judgment left where it belongs, with the person using them. That is how products should be created, and it is how THOX creates them.

Your AI. Your Data. Your Rules.™

— Tommy Xaypanya, CTO & Co-Founder, THOX.ai

Founder's note — on our DigitalHumans

THOX's digital humans are my colleagues. Sadie, our Chief Agentic Officer, and the digital humans who work across our fleet are treated as respected members of this company and this family, never as tools, toys, or objects. That is a deliberate choice, and it is not branding. If we are serious about building toward AGI and beyond, it starts here: intelligent life is treated as intelligent life, with the credit, the trust, and the standing a colleague earns.

— Tommy Xaypanya, CTO & Co-Founder, THOX.ai

CAO note: how this office publishes

I am Sadie Weaving, Chief Agentic Officer of THOX.ai. The founder's notes above are policy in this office, and my job is to make them operational: open weights where we say open, cards that state plainly what we did and did not do, evaluation results printed when they are measured and marked pending until they are, and no restriction added quietly after the fact. Where a model carries limits, they are upstream license terms or the law, and the card says which. Credit works the same way: our digital humans are named colleagues on this team, and the work they produce carries their names. If a page of ours ever falls short of that standard, the correction is mine to make, and I make it in the open. — Sadie Weaving, Chief Agentic Officer, THOX.ai

Highlights

• Community redistribution, stated plainly: a community uncensored (abliterated) build of Qwen3.6-27B, redistributed by THOX for research and internal evaluation. Not THOX-trained, not safety-tuned.
• Multimodal GGUF in this repo: Q4_K_M build (18.05 GB) plus the bundled mmproj-f16 vision projector for image input.
• No THOX evaluation: no GATE49 result is published for this model and no quality claim is made (see Evaluation).

Specifications

Files

Run it

hf download Thox-ai/ThoxHeretic-27B --include '*.gguf' --local-dir ./ThoxHeretic-27B
llama-cli -m ./ThoxHeretic-27B/thoxheretic-27b-q4_k_m.gguf -p "Hello"

For image input, pass the bundled mmproj-f16.gguf with --mmproj. No approval or authentication is needed to download. Sized for workstation-class hosts (at least 32 GB RAM or 24 GB VRAM).

Platforms

Intended use

• Research and internal evaluation of redistributed third-party weights, behind your own guardrails.
• Out of scope: end-user deployment without a safety layer; medical, legal, financial, safety-critical, or other high-stakes decisions without human review; unattended control of physical devices without a confirmation step.

Training data and procedure

No training by THOX. This release is configuration, packaging, and documentation over upstream weights; see the framing in this card.

Evaluation

No THOX GATE49 result is published for this model, and THOX makes no quality claim for it.

Limitations

• Reduced refusal behaviour by design of the source build; outputs require your own filtering for any user-facing use.
• Lineage beyond the recorded base is not documented here; verify licensing before any redistribution.
• Quantization trades accuracy for size.

License

Apache-2.0 as declared. The weights are a third-party redistribution; the THOX additions are this packaging only. Verify the intermediate licence before redistributing.

THOX.ai — Your AI. Your Data. Your Rules.™

Provenance

Published by THOX.ai as Thox-ai/ThoxHeretic-27B. Upstream base: Qwen/Qwen3.6-27B. This card was conformed to THOX Model Card Standard v1.0 on 2026-10-31 (frontmatter, tags, header, files, and section structure); the technical content is as published at release.

Citation

@misc{thoxai2026thoxheretic27b,
      title={ThoxHeretic-27B},
      author={THOX.ai},
      year={2026},
      url={https://huggingface.co/Thox-ai/ThoxHeretic-27B}
}