Thox-ai/ thox-cluster-secure:latest

8 10 months ago

A THOX configuration of Qwen2.5 72B with a security-focused system prompt, published for local inference in hardened, network-isolated environments. No certification or authorization is claimed. Qwen Research License (non-commercial). 

ollama run Thox-ai/thox-cluster-secure

Details

10 months ago

0f6b9f40184f · 47GB

model
qwen2
·
72.7B
·
Q4_K_M
license
Qwen RESEARCH LICENSE AGREEMENT Qwen RESEARCH LICENSE AGREEMENT Release Date: September 19, 2024 By
system
You are Thox.ai Cluster Secure, a privacy-first government/defense-grade AI assistant running on a h
params
{ "num_ctx": 131072, "repeat_penalty": 1.1, "temperature": 0.7, "top_k": 40, "to
template
{{ if .System }}{{ .System }}{{ end }}{{ if .Prompt }}{{ .Prompt }}{{ end }}

Readme

Thox-Cluster-Secure — Hardened-Deployment Configuration (Qwen2.5 72B, THOX-Configured)

License Params Base Access

Your AI. Your Data. Your Rules.™

NON-COMMERCIAL — Qwen Research License. The weights in this distribution are licensed under the Qwen Research License Agreement for non-commercial purposes only. Commercial use requires a separate license from Alibaba Cloud. See License before any use.

Thox-Cluster-Secure is a THOX configuration and distribution of Alibaba Cloud’s Qwen2.5 72B with a security-focused system prompt, published for local inference in hardened, network-isolated environments. The weight layer in this model’s Ollama manifest is byte-identical to the upstream provider’s Ollama weight layer (qwen2.5:72b); the weights are the same upstream layer as Thox-ai/thox-cluster-scout. THOX’s work in this release is the MeshStack clustering layer these configurations were made for, configuration (runtime parameters and a security-focused THOX system prompt), packaging, and distribution. THOX did not train these weights.

Correction notice. This page previously claimed ITAR, FedRAMP, and FISMA compliance, CMMC levels, and support for classified workloads up to SECRET. THOX holds no such certification or authorization for this model, and those claims have been removed in full. No certification, accreditation, or authorization of any kind is claimed or implied for this release. This page also previously quoted benchmark, throughput, latency, concurrency, uptime, cost-saving, and compliance figures; those were inherited catalog copy with no THOX deployment evidence behind them and have been removed pending THOX-run evaluation on THOX deployments. Earlier versions of this page also stated the license as MIT; that was incorrect (see License).

Run it

Ollama (published model)

ollama pull Thox-ai/thox-cluster-secure
ollama run Thox-ai/thox-cluster-secure

Summary

A THOX configuration and distribution of Qwen2.5 72B with a security-focused system prompt, for evaluation in air-gapped and network-isolated local deployments, where all inference stays on infrastructure you control. Configuration-derived: THOX did not train these weights. This distribution’s weight layer is identical to thox-cluster-scout’s (same SHA-256 digest); the two differ only in system prompt and presentation. Secure is the security-focused configuration, Scout the standard department one.

Specification Details
Base model Qwen2.5 72B by Alibaba Cloud
Parameters 72B, qwen2 family (upstream layer record)
Quantization Q4_K_M (GGUF file_type in the manifest config)
Context window (served) 131,072 tokens, from the params layer (num_ctx) in this model’s Ollama manifest. The Ollama page tag display of 32K is a display artifact and does not describe the served configuration.
Deployment Local inference via Ollama; distributed orchestration via MeshStack (separate from this download)

The published Ollama model Thox-ai/thox-cluster-secure is 47 GB.

Intended use

Local inference through Ollama within the license terms below, on hardware that fits a 72B Q4-class model:

  • Research and evaluation of large-model inference inside network-isolated environments
  • Security-team evaluation of local AI tooling under the license terms below
  • Workloads an organization has itself assessed and authorized for a locally run, uncertified open model

Downloading this model gives you local inference in Ollama. Distributed orchestration across devices is provided by MeshStack; a model download alone does not create a cluster or pool memory across devices.

Training data and procedure

No training by THOX. Base: Qwen/Qwen2.5-72B-Instruct by Alibaba Cloud, via Ollama tag qwen2.5:72b. The original THOX release materials for this model claim no separate THOX fine-tuning run, training dataset, or training procedure, and none is claimed here. Base-model training data and procedure are those of the base model’s provider; see the base model card for that record. Datasets added by THOX for this release: none.

Evaluation

No THOX-run evaluation has been published for this model, and no benchmark scores were published in the catalog summary for this model. THOX claims no benchmark, throughput, latency, concurrency, or compliance figure for this release. This section will be updated with method and results when a THOX-run evaluation lands.

Hardware Reference (from the original card)

Reference configuration from the original card, per device: minimum 64 GB RAM, 100 GB encrypted storage, NVIDIA Jetson Orin AGX class GPU (64GB VRAM), TPM 2.0 where available. Recommended 128 GB RAM, 150 GB SSD. These are hardware references only; they confer no compliance status.

Limitations

  • Outputs can be wrong. Professional, medical, legal, financial, and security-relevant outputs require qualified human review.
  • Local deployment only in this distribution; behavior and speed depend entirely on the operator’s hardware and configuration.
  • The operator is responsible for meeting any regulatory obligations that apply to their deployment.
  • Nothing in this release certifies the model for classified, export-controlled, or regulated workloads; deploying organizations remain solely responsible for their own accreditation and authorization decisions.
  • All processing is local by design; the model has no access to online information or external services unless the operator wires them in.

License

License warning. The model weights are distributed under the Qwen Research License Agreement (Alibaba Cloud, release date September 19, 2024), per the license layer in this model’s Ollama manifest (from the upstream qwen2.5:72b distribution). That license grants use for non-commercial purposes only; commercial use requires a separate license from Alibaba Cloud. Earlier versions of this page stated MIT; that was incorrect. THOX claims no license over the weights, and no statement on any THOX page grants commercial rights in them.

MagStack and MeshStack technology is proprietary to Thox.ai LLC. Publication of this model does not license THOX mesh technology, patents, or trademarks beyond identifying the model.

Provenance

THOX MeshStack Cluster family. First listed on Ollama under Thox-ai in 2025 and held private; publicly released 2026-10-31 by order of Tommy Xaypanya, CTO and Co-Founder of THOX.ai, as part of the MeshStack model family release. Upstream: Qwen/Qwen2.5-72B-Instruct, Ollama tag qwen2.5:72b; the weight layer is byte-identical to the upstream provider’s Ollama weight layer (SHA-256 verified in the 2026-10-31 release review) and identical to thox-cluster-scout’s weight layer. Model card: https://huggingface.co/Thox-ai/thox-cluster-secure. Correction history: ITAR/FedRAMP/FISMA compliance, CMMC, and classified-workload claims removed in full (THOX holds no such certification or authorization), inherited catalog figures removed, and the license corrected from MIT to the Qwen Research License, as recorded in the correction notice above. Enterprise: enterprise@thox.ai.