Compressed decision models
Third-party decision models in smaller GGUF builds from TurinTech, built with Artemis. See each card for tests and hardware.
Text Classification • 9B • Updated • 703 • 9Note Start here. Cloudflare Clef-Flash 9B. The default 4.9 GB file gives the same answer as the original on 95.0% of 2,000 test decisions (CPU). Text and images. Peak GPU memory 5.5 GiB; the public 6.5 GB Q4_K_M is a little closer.
turintech/Clef-Artemis-GGUF
Text Classification • 27B • Updated • 696 • 9Note Cloudflare Clef 27B. The default 14.8 GB file gives the same answer as the original on 95.6% of 2,000 test decisions. Text and images. Peak GPU memory 15.0 GiB for text; images need about 2.4 GiB more.
turintech/pplx-decider-27B-Artemis-GGUF
Text Classification • 27B • Updated • 727 • 9Note Perplexity Decider 27B. The default 15.3 GB file gives the same answer as the original on 96.8% of 2,000 test decisions. Text and images. Peak GPU memory 16.6 GiB, so plan on a 24 GB GPU.
turintech/Kev-9B-Artemis-GGUF
Text Classification • 9B • Updated • 755 • 8Note Kev-9B by Jared Palmer. The default 4.8 GB file gives the same answer as the original on 92.2% of 2,000 test decisions. Text only. Peak GPU memory 9.1 GiB; the public 6.4 GB Q4_K_M is closer (93.2%).
turintech/Kev-4B-Artemis-GGUF
Text Classification • 4B • Updated • 672 • 8Note Kev-4B by Jared Palmer. The default 2.5 GB file gives the same answer as the original on 93.0% of 2,000 test decisions. Text only. Peak GPU memory 7.6 GiB; 1.7 points less accurate on held-out decisions.
-
turintech/d3-flash-Artemis-GGUF
Text Classification • 9B • Updated • 459 • 4 -
turintech/d3-27B-Artemis-GGUF
Text Classification • 27B • Updated • 870 • 4 -
turintech/Kev-27B-Artemis-GGUF
Text Classification • 26B • Updated • 473 • 1