Organisation card: Artemis Edge collections

README.md

@@ -1,10 +1,32 @@
 ---
 title: README
-emoji: 🐢
-colorFrom: purple
-colorTo: purple
+emoji: 🎯
+colorFrom: indigo
+colorTo: green
 sdk: static
 pinned: false
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# TurinTech
+
+TurinTech builds Artemis, a code optimisation platform that searches for code changes and tests them against a project's own benchmarks. Here we publish smaller GGUF versions of open models.
+
+## Artemis Edge
+
+Our five decision models and two Gemma releases were compressed by TurinTech with techniques discovered by Artemis.
+
+**[Decision models](https://huggingface.co/collections/turintech/artemis-edge-decision-models-6accea74eb7016a886f25fdb)**: Cloudflare's Clef 27B and Clef-Flash 9B, Perplexity's Decider 27B, and Jared Palmer's Kev-9B and Kev-4B. Give them text and a question with a fixed set of answers, and they return a probability for each answer through llama.cpp's `/v1/systemone` endpoint. Clef, Clef-Flash and the Decider also take images, with the projectors named in their cards.
+
+**[Compressed LLMs](https://huggingface.co/collections/turintech/artemis-edge-compressed-llms-6acd251457ba2b9f33ad9d8a)**: text-only builds of Google's Gemma 4 E4B and E2B, plus our earlier GLM-5.2 MaxMin108 release, which runs on a 128 GB GB10 desktop with the quality trade-offs listed in its card.
+
+The decision-model cards report how often each file gives the same answer as its original, the accuracy change, and whether any answer the original was more than 80% sure of changed. The Gemma cards report next-token agreement, short continuations and task accuracy. Each card also compares our files with public files of a similar size, including where a public file does better.
+
+The models run on stock llama.cpp. Each card lists the version, settings and companion files they need.
+
+These are independent TurinTech releases, not affiliated with or endorsed by the original model authors. Each card links to its sources, credits and licence.
+
+## Custom versions
+
+We fine-tune models on an organisation's own data and build versions for its hardware, including CPU servers, laptops, Apple Silicon and edge devices.
+
+[Book a demo](https://www.turintech.ai/book-a-demo) · [turintech.ai](https://www.turintech.ai)