d3 27B decision model in GGUF, compressed by TurinTech with Artemis. Use the /v1/systemone endpoint, not chat. Recommended: d3-27B-Artemis-IQ4_XS.gguf. Model files, checksums and measured trade-offs: https://huggingface.co/turintech/d3-27B-Artemis-GGUF. Download size is not runtime memory. Measured IQ4_XS: same answer as BF16 97.1%, 0 confident answers changed on LocalLLaMA/typed-decisions test split, typed decisions; protocol rtx5090-fcac190-20261014.