Introducing our Artifacts Hub and Adoption Dashboard

Interconnects has quietly turned its side-project data hoarding into something the rest of us can actually use. The site just launched two standalone tools: an Artifacts Hub cataloging 792 open models released over the past two years, and an Adoption Dashboard tracking download and derivative-model numbers by geography and organization. Both are free, both update regularly, and both grew out of work the team was already doing for internal projects like The ATOM Project.

The Artifacts Hub is the deeper of the two. Built in collaboration with Project VAIL, an AI verification startup, it pulls in Hugging Face trending data, inference token counts from Open Router, and intelligence benchmarks from Artificial Analysis. For a model like GLM-5.2, you can see at a glance how far it trails the frontier on Artificial Analysis's Intelligence Index, how its Hugging Face and Open Router traction stacks up against comparable models, and something Interconnects calls a Relative Adoption Metric — essentially downloads normalized for model size and time since release. There's also a VAIL similarity score for spotting related model families. It's a curated slice of a much bigger dataset; Interconnects says it tracks every model on Hugging Face, crunches a public list of a few thousand core LLMs on GitHub, and hand-picks a few hundred of those for this closer look.

The Adoption Dashboard is simpler but has been a long time coming. Ever since publishing The ATOM Project, the team kept getting asked to update its US-versus-China open-model adoption chart, and doing that by hand got old fast. Now the underlying numbers refresh daily, showing not just the US-China gap but which other countries and organizations are gaining ground in open-weight releases.

Neither project is framed as a product launch so much as an infrastructure drop for people trying to figure out whether open models are actually catching on outside labs and leaderboards. Interconnects says the goal is transparency — helping the ecosystem see what's working as companies try to make open models genuinely cost-competitive with closed frontier ones. They're also floating the data for outside research or products, and asking for feedback directly by email, which is a refreshingly low-tech way to end a data-dashboard announcement.