{"slug": "i-built-a-registry-for-system-one-models-here-s-what-i-learned-comparing-all-of", "title": "I built a registry for System One models — here's what I learned comparing all of them", "summary": "A developer built systemonemodels.tech, a registry cataloguing more than fifteen System One model families that read a state and answer typed questions — multiple choice, scale scores, or probabilities — in a single forward pass without generating text. The registry lists each model's maker, size, context, licence, price and reported numbers, and highlights that reported accuracies are not comparable across publishers' differing test suites, that small encoders such as the 421M-parameter Laya can beat general models on narrow decisions when fine-tuned on local labels, and that many entries are fine-tunes of fine-tunes. The project ships a CLI (pip install systemonemodels) for pulling models and a local studio for fine-tuning.", "body_md": "*Disclosure: I'm the founder of systemonemodels.tech, the registry this post is about.*\n\nTen days after TypeSafe AI shipped Jev and named the category, there were more than fifteen System One model families. A System One model reads a state and answers typed questions — choose an option, score on a scale, or give the probability that a condition holds — in one forward pass, with calibrated confidence and no generated text.\n\nThey were scattered across Hugging Face repos, blog posts and API docs, each reporting accuracy on its own test set. So I built one place for them: [systemonemodels.tech](https://systemonemodels.tech/) lists every one with its maker, size, context, licence, price and reported numbers.\n\nThree things stood out:\n\n**The numbers don't rank anything yet**. Every publisher reports on a different suite. The only accuracy that means something is the one you measure on your own labelled examples.\n\n**Small encoders are competitive for narrow decisions**. Laya is a 421M-parameter ModernBERT encoder with a decision head, and it answers a question in tens of milliseconds on a GPU. Fine-tuned on your own labels, it can beat a general model on your decision.\n\n**Lineage matters**. Many models are fine-tunes of fine-tunes. Each registry page shows the full model tree, back to the pretrained backbone.\n\nIf you want to try one:\n\n`pip install systemonemodels`\n\nsystemone pull convai-innovations/laya\n\nAnd to fine-tune on your own data, locally:\n\n`systemone run studio`\n\nEverything is free to browse and publish. Feedback welcome — especially on what a fair benchmark across these models should look like.", "url": "https://wpnews.pro/news/i-built-a-registry-for-system-one-models-here-s-what-i-learned-comparing-all-of", "canonical_source": "https://dev.to/biplov_gautam/i-built-a-registry-for-system-one-models-heres-what-i-learned-comparing-all-of-them-25nh", "published_at": "2026-09-27 15:00:49+00:00", "updated_at": "2026-09-27 15:31:12.694238+00:00", "lang": "en", "topics": ["ai-tools", "machine-learning", "large-language-models", "ai-products"], "entities": ["systemonemodels.tech", "TypeSafe AI", "Jev", "Laya", "ModernBERT", "Hugging Face", "convai-innovations"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/i-built-a-registry-for-system-one-models-here-s-what-i-learned-comparing-all-of", "markdown": "https://wpnews.pro/news/i-built-a-registry-for-system-one-models-here-s-what-i-learned-comparing-all-of.md", "text": "https://wpnews.pro/news/i-built-a-registry-for-system-one-models-here-s-what-i-learned-comparing-all-of.txt", "jsonld": "https://wpnews.pro/news/i-built-a-registry-for-system-one-models-here-s-what-i-learned-comparing-all-of.jsonld"}}