ThoxNova-12B-Core

Conservative Chat LoRA — THOX’s conservative chat fine-tune for the THOX stack: a LoRA adapter (rank 8), bundled with Q4_K_M (7.38 GB) and Q5_K_M (8.55 GB) GGUF builds

Safetensors weights · GGUF: Q4_K_M, Q5_K_M · Live on Ollama

Your AI. Your Data. Your Rules.™

Status: PUBLISHED. This repository carries the LoRA adapter (safetensors) and the bundled Q4_K_M and Q5_K_M GGUF builds — the full runnable builds — and the model is live on Ollama as Thox-ai/thoxnova-12b-core. See Files for exactly what this repo contains.

ThoxNova-12B-Core is THOX’s conservative chat fine-tune for the THOX stack: a rank-8 LoRA adapter on Gemma-4-12B-IT, bundled with Q4_K_M and Q5_K_M GGUF builds so the runnable model ships in this one repository. THOX ran the fine-tune and the GGUF conversions; the base model is Google’s. The same THOX adapter is also published as ThoxNova-12B-Agent with a different persona overlay. This repository is public and ungated, and this card is published together with the weights.

Summary

ThoxNova-12B-Core is THOX’s conservative chat fine-tune for the THOX stack: a LoRA adapter (rank 8), bundled with Q4_K_M (7.38 GB) and Q5_K_M (8.55 GB) GGUF builds. The same THOX adapter is also published as ThoxNova-12B-Agent with a different persona overlay. Requires a llama.cpp or Ollama build with support for this model family.

Chat template:

<start_of_turn>user
{prompt}<end_of_turn>
<start_of_turn>model

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

• Adapter plus two runnable builds in one repo: the rank-8 LoRA adapter (safetensors) with bundled Q4_K_M (7.38 GB) and Q5_K_M (8.55 GB) GGUF builds.
• Conservative chat posture: the Core persona overlay for on-device and edge chat in the THOX stack; the same adapter ships as ThoxNova-12B-Agent with a different persona overlay.
• Two load paths: attach the adapter to the base model with Transformers + PEFT, or run a bundled GGUF through Ollama or llama.cpp.

Specifications

Files

Run it

Bundled GGUF (Ollama)

ollama run Thox-ai/thoxnova-12b-core

Bundled GGUF (llama.cpp)

hf download Thox-ai/ThoxNova-12B-Core --include '*.gguf' --local-dir ./ThoxNova-12B-Core
llama-cli -m ./ThoxNova-12B-Core/thoxnova-12b-core-Q4_K_M.gguf -p "Hello" -n 128

Adapter (Transformers + PEFT)

Load the base model named in this card’s metadata, then attach the adapter from this repo:

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
base = "<base model named in this card's metadata>"
tok = AutoTokenizer.from_pretrained(base)
model = AutoModelForCausalLM.from_pretrained(base, device_map="auto")
model = PeftModel.from_pretrained(model, "Thox-ai/ThoxNova-12B-Core")

Sized for hosts with at least 16 GB RAM or a GPU with at least 10 GB VRAM.

Platforms

Use with ThoxOS

ThoxOS loads this model as a host-class assistant through its local model runtime. Serve the bundled GGUF with Ollama on the ThoxOS host:

ollama run Thox-ai/thoxnova-12b-core

ThoxOS points its chat and agent surfaces at the local Ollama endpoint (http://localhost:11434); no cloud account is required and prompts stay on the machine.

Use with MeshStack

On a MeshStack node, pull the model once and serve it to the rest of your mesh:

ollama pull Thox-ai/thoxnova-12b-core
ollama serve

Other devices on the same mesh reach it through the node’s OpenAI-compatible endpoint (http://<node>:11434/v1). The adapter form can equally be served from the Hugging Face files with llama.cpp:

hf download Thox-ai/ThoxNova-12B-Core --include '*.gguf' --local-dir ./ThoxNova-12B-Core
llama-cli -m ./ThoxNova-12B-Core/thoxnova-12b-core-Q4_K_M.gguf -p "Hello" -n 128

Intended use

• On-device and edge chat within the THOX stack.
• Not a safety-aligned public assistant unless deployed behind THOX guardrails.
• Out of scope: 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

THOX fine-tune of google/gemma-4-12b-it. The corpus and procedure are described in this card as published at release (see Summary and Evaluation); no training beyond what this card describes is claimed for this repository.

Evaluation

No THOX GATE49 result is published for this model.

Limitations

• No evaluation artifact or training log is published for this model; no quality figure is claimed.
• Models of this size make factual mistakes and can produce confident but wrong output; verify important answers.
• Behaviour and bias reflect the training data, which is mostly English; quality in other languages is lower.
• Quantization trades accuracy for size.

License

Apache-2.0 as declared (see the linked license terms). The THOX adapter and additions in this repo are covered under Apache-2.0; keep the upstream LICENSE/NOTICE with any redistribution.

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

Provenance

Published by THOX.ai as Thox-ai/ThoxNova-12B-Core. Upstream base: google/gemma-4-12b-it. 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{thoxai2026thoxnova12bcore,
      title={ThoxNova-12B-Core},
      author={THOX.ai},
      year={2026},
      url={https://huggingface.co/Thox-ai/ThoxNova-12B-Core}
}