ThoxRoute-1.5B-Service-GGUF

Service Router GGUF — a single Q4_K_M GGUF intended for low-memory MeshStack and ThoxOS service nodes

GGUF: thoxroute-service-q4_K_M.gguf

Your AI. Your Data. Your Rules.™

Status: UNVALIDATED ARTIFACT. No runtime validation has been performed or published for this artifact: it has not been loaded, benchmarked, or functionally checked, and nothing on this card is evidence that it routes correctly. The card states this in full under Evaluation.

Summary

A single Q4_K_M GGUF intended for low-memory MeshStack and ThoxOS service nodes. Licence resolved as Apache-2.0 on 2026-10-31 on the corroborated lineage set out below, with the residual caveat printed in full. This card is not a certification, a benchmark, or a shipping approval. Everything stated about the file itself is read from the Hugging Face Hub's own record of this repository or from the artifact's own GGUF header.

Highlights

• A single Q4_K_M GGUF intended for low-memory MeshStack and ThoxOS service nodes.

Specifications

Files

The artifact

Read from the GGUF header by the Hub's metadata parser: architecture qwen2, context length 32768, tensor total 1,777,088,000, BOS <|endoftext|>, EOS <|im_end|>. The embedded chat template is the stock instruct template, including its default upstream system persona; it does not carry a THOX routing template. Unless your host overrides the system prompt, the embedded template will introduce the model under its upstream identity.

Provenance — corroborated, with a printed caveat

The repository's commit history states the intent was a quantization of Thox-ai/ThoxRoute-1.5B. On 2026-10-31 the artifact's own GGUF header was read directly and corroborates that lineage: the embedded general.name is "Service v1 Merged" (a merged build, as a service deployment of the adapter program would be), general.architecture is qwen2, and general.size_label is 1.8B, matching the Qwen2.5-1.5B base of the stated source. THOX fleet records independently show this exact repository deployed as the ThoxRoute classifier service model.

Caveat, printed in full: no build record, conversion log, or hash chain survives that links this file to that adapter byte-for-byte, and one discrepancy remains unexplained: the sibling repository Thox-ai/ThoxRoute-1.5B-GGUF holds its Q4_K_M at 986,047,936 bytes; this file is about 131 MB larger for the same nominal quantization of the same nominal source. The licence resolution below rests on the corroboration above plus the stated lineage, and was accepted on that basis by Tommy Xaypanya, CTO, THOX.ai, on 2026-10-31.

Licence — Apache-2.0 (resolved 2026-10-31)

This repository is published under the Apache License 2.0. Basis: the stated source, Thox-ai/ThoxRoute-1.5B, is published under Apache-2.0, and its base, Qwen/Qwen2.5-1.5B-Instruct, is also Apache-2.0, so a merge and quantization of the two carries Apache-2.0. The corroboration and the residual caveat behind this determination are printed in the Provenance section above; the determination was made by Tommy Xaypanya, CTO, THOX.ai, on 2026-10-31.

Run it

hf download Thox-ai/ThoxRoute-1.5B-Service-GGUF thoxroute-service-q4_K_M.gguf --local-dir .
sha256sum thoxroute-service-q4_K_M.gguf
# expect dcb4d041fa2ae90b5c3b713e9bd8e6c4b265388f4f81c804ae7a404737c4c1b2
llama-cli -m thoxroute-service-q4_K_M.gguf -p "Route: reset my password"

ollama run hf.co/Thox-ai/ThoxRoute-1.5B-Service-GGUF

There is no published ollama.com/thox-ai/thoxroute-service tag; ollama.com/thox-ai/thoxroute serves the standard router. Use the hf.co/ form above, or the standard router tag.

Platforms

Intended use

• On-device intent and preference routing inside the THOX stack, on service nodes where the standard routing GGUF does not fit.
• Deploy behind THOX guardrails on any public-facing path. Not a general assistant; not safety-aligned for direct public exposure.

Training data and procedure

THOX fine-tune of Qwen/Qwen2.5-1.5B-Instruct. 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, and no runtime validation has been performed or published for this artifact: it has not been loaded, benchmarked, or functionally checked as part of this card. Nothing here is evidence that it routes correctly.

Limitations

• The licence determination rests on corroborated lineage, not a surviving hash chain (see Provenance). That is the primary limitation.
• Quantization to Q4_K_M can change routing decisions at the margins, and no comparison against the unquantized adapter has been run.
• No throughput, accuracy, benchmark, compliance, or cost figures are claimed.
• The embedded chat template is the stock upstream template, not a router template (see above).
• A successful download is not the same as an inference-ready service: verify the SHA-256 and run your own smoke test.

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

License

Licensed under apache-2.0 as declared in this card's metadata; the license text in this repository governs.

Citation

@misc{thoxai2026thoxroute15bservicegguf,
      title={ThoxRoute-1.5B-Service-GGUF},
      author={THOX.ai},
      year={2026},
      url={https://huggingface.co/Thox-ai/ThoxRoute-1.5B-Service-GGUF}
}