--- pretty_name: "Qwen3.8-27B on 12 GB — frozen deployment measurements" license: other tags: - reproducibility - llm-inference - quantization - consumer-gpu configs: - config_name: summary data_files: - split: train path: data/summary.jsonl - config_name: throughput data_files: - split: train path: data/throughput/phase1.jsonl --- # Qwen3.8-27B on 12 GB — result dataset - Matthew Schwartz — [ORCID 0009-0009-4171-7247](https://orcid.org/0009-0009-4171-7247) This dataset mirrors the redistribution-cleared measurements for “Deploying Qwen3.8-27B in 12 GB of VRAM: Accuracy and Throughput Across Quantized Inference Stacks.” This is the 1.0.1 correction, dated 2026-09-04. Original measurement rows are unchanged; see [CORRECTIONS.md](CORRECTIONS.md). The previous 1.0.0 artifact remains at [doi:10.5281/zenodo.22166977](https://doi.org/10.5281/zenodo.22166977). The correction's version DOI is [doi:10.5281/zenodo.22314395](https://doi.org/10.5281/zenodo.22314395). Version-specific source and analysis instructions are at . The result dataset is . In prepublication review copies, the reserved DOI and planned tag need not resolve until release; this card alone is not a publication receipt. This repository contains measurement records under mixed terms rather than a single blanket license. Original measurements and documentation are CC BY 4.0; benchmark-derived fields retain their upstream terms. See `LICENSE` and `THIRD_PARTY_NOTICES.md`. The study measured eight complete artifact–engine deployments on one 80 W RTX 4080 Laptop GPU. It is a single-machine characterization, not a universal model or quantization ranking. Full model outputs were not retained; the per-item files contain scoring fields and short answer excerpts. No model weights are included. Exact upstream weight revisions and hashes are in `environment/model_artifacts.json`. The full analysis-only verifier and fixed benchmark subsets live in the canonical GitHub/Zenodo artifact rather than this discovery mirror. From that artifact root, run: ```bash python3 check_claims.py ``` ## AI assistance Claude (Anthropic) and Codex (OpenAI) assisted with evaluation-harness development, experiment execution, evidence auditing, analysis and figure code, and manuscript drafting and revision. Matthew Schwartz directed the work and reviewed and verified the retained outputs against the archived evidence and cited sources, edited the manuscript, and takes full responsibility for the methods, results, interpretation, citations, and released artifacts.