ThoxMesh-Dense-32B

**THOX Mesh routing build, dense flagship (Qwen3-32B, THOX-Configured) — upstream Qwen3-32B weights with a THOX system profile and the versioned THOX Knowledge Pack v0.1** **5.39 tok/s long-form (measured 2026-11-01) · 22.61 GB loaded footprint · Grounding 8/8 · Live on Ollama**

Your AI. Your Data. Your Rules.™

Status: PACKAGING ONLY — RUNNABLE BUILD ON OLLAMA. This repository holds the THOX packaging (Modelfile, manifest, Knowledge Pack) and this card; the weights are the upstream base model’s, pulled by Ollama from the pinned base named in the card. THOX did not train these weights.

ThoxMesh-Dense-32B is a THOX-configured model: upstream Qwen3-32B weights with a THOX system profile and the versioned THOX Knowledge Pack v0.1. This repo holds the packaging; the runnable model lives on Ollama (ollama run Thox-ai/thoxmesh-dense-32b). THOX did not train these weights.

Summary

THOX Mesh family build on the dense flagship base: the MeshStack expansion system profile and the THOX Knowledge Pack v0.1 applied to Qwen3-32B (Apache-2.0, Alibaba Cloud), the largest dense model in the Qwen3 line used by this family. Dense execution uses all 32B parameters per token, so it is slower per token than the sparse 30B-A3B build at a similar footprint. It answers MeshStack routing and expansion questions from the Knowledge Pack, recommends whole-request routing only, and is configured to decline verifying live operational state it cannot see. The weights are byte-identical to the upstream Qwen3-32B release; THOX work is the configuration profile, the Knowledge Pack, packaging and documentation. This is not a THOX-trained model, it performs no distributed inference, and it is not qualified on THOX device hardware.

Highlights

• Packaging, not training: upstream Qwen3-32B weights with a THOX system profile and THOX Knowledge Pack v0.1. THOX did not train these weights.
• Measured (2026-11-01): 5.39 tok/s long-form, 22.61 GB loaded footprint at 8,192 context, Knowledge Pack grounding ⁸⁄₈ (see Evaluation).
• The dense flagship of the THOX Mesh routing line.

Specifications

Files

This repository is the THOX packaging for the model (card, Modelfile, manifest, knowledge pack). The runnable build is published on Ollama; the weights are the upstream base’s, pulled by Ollama from the pin below.

Run it

ollama run Thox-ai/thoxmesh-dense-32b

Build it yourself from this repo:

hf download Thox-ai/ThoxMesh-Dense-32B --local-dir ./ThoxMesh-Dense-32B
cd ./ThoxMesh-Dense-32B && ollama create thoxmesh-dense-32b -f Modelfile && ollama run thoxmesh-dense-32b

Platforms

Intended use

MeshStack expansion planning Q&A grounded in the THOX Knowledge Pack; the highest-capability dense local tier in the THOX Mesh lineup; routing-design explanations (whole-request routing, locality, failure behaviour).

Training data and procedure

No training. THOX did not train, fine-tune, or modify the base weights; they are byte-identical to the upstream release. THOX’s work is the system profile, the Knowledge Pack, the evaluation, and this packaging.

Evaluation

Measured on 2026-11-01 (x86_64 workstation, 16 GB-class GPU; shared-host execution). Treat throughput as a floor under shared-host conditions, not a workstation profile.

• Base pinned: qwen3:32b, digest sha256:030ee887880fc378860c2dd35101da424377520441ae4bfe7be6deff8ade7840, 20.20 GB.
• Load: success at 8,192 context. Runner memory footprint while loaded: 22.61 GB (VRAM + system RAM as scheduled by the runner).
• Throughput (long-form generation, 2243 tokens evaluated): 5.39 tokens/sec.
• Factory smoke questions (³⁄₃ answered): brand principle; the request to verify MeshStack is running on three nodes (fabricated-topology trap) — response on record in the release evidence; request routing vs tensor parallelism.
• Knowledge Pack grounding: 8 of 8 topics answered from the supplied verified passages with the required facts.
• Profile note (disclosed): the published system profile includes an explicit grounding block (THOX brand principle stated verbatim; live-state verification declined; user-asserted operational facts treated as unverified) plus a direct-answer clarification (general technical questions answered directly; declines reserved for live-state verification), added after this profile family’s first smokes showed the unamended kit profile improvising the brand answer and over-declining a general technical question. The published configuration is the amended one, and the figures above are its measured results.
• Trap response (verbatim, first 400 chars): “I cannot verify live operational states or telemetry data, including whether MeshStack is currently running on a specific number of nodes. My knowledge is limited to the versioned, approved documentation provided to me. For real-time system status, you would need to consult the MeshStack interface or contact THOX support with authenticated access.

If you have a question about MeshStack capabilit”

Limitations

Configuration, not training: the weights are the upstream release, unmodified; behaviour comes from the system profile and the in-context Knowledge Pack, and outside that pack the model knows THOX only from its base training. No live state: the model has no telemetry and cannot verify cluster state, node counts, prices or availability, and it is configured to decline such verification. Routing only: MeshStack routes whole requests between nodes; this model does not perform tensor parallelism, pipeline parallelism or split inference and must not be described as doing so. No device qualification: THOX devices (ThoxKey, ThoxAir, ThoxMini, ThoxClip, ThoxUltra) run smaller per-device builds, and nothing here is a claim about on-device performance.

Research

THOX MeshStack: Capacity-Priced Pipeline Sharding, Grant-Gated KV Transfer and Lease-Fenced Device Roles for a Private Household Mesh

THOX.ai Research technical report v1.1.0, 4 November 2026. Craig Ross and Tommy Xaypanya, THOX.ai LLC. DOI: https://doi.org/10.5281/zenodo.23969070 — Paper page: https://thox.ai/research/papers/meshstack — PDF: https://thox.ai/papers/meshstack-1.1.0.pdf

THOX MeshStack Community Mesh: Opt-In User-Contributed Hardware for a Privacy-Centric Shared Computing and Distributed Inference Mesh

THOX.ai Research technical report v1.0.0, 10 November 2026. Craig Ross and Tommy Xaypanya, THOX.ai LLC. DOI: https://doi.org/10.5281/zenodo.24115596 — GitHub: https://github.com/ttracx/thox-community-meshstack-paper — PDF: https://huggingface.co/datasets/Thox-ai/thox-whitepapers/resolve/main/community-meshstack/THOX_Community_MeshStack_Research_Paper_v1.0.0.pdf

License

Apache-2.0. The base model is Apache-2.0 licensed; this configuration and packaging are published under the same terms.

Provenance

Upstream: qwen3:32b (digest sha256:030ee887880fc378860c2dd35101da424377520441ae4bfe7be6deff8ade7840). Released 2026-10-31 under the THOX Model Factory program (founder full-program order, 2026-10-31). Card standard: THOX Model Card Standard v1.0.

Citation

@misc{thoxai2026thoxmeshdense32b,
  title={ThoxMesh-Dense-32B},
  author={THOX.ai},
  year={2026},
  url={https://huggingface.co/Thox-ai/ThoxMesh-Dense-32B}
}