Open-weight models
AI models whose trained parameters are published for anyone to download, run, and modify — as distinct from models accessible only through an API.
A language model is a file of numbers — the weights — that encode what it learned from training. Open-weight means those numbers are published: anyone can download the model, run it on their own hardware, and fine-tune it for their own purpose. That distinguishes them from closed models, where the weights stay on the provider's servers and you interact only through an API. The hard part is the gap: the most capable models remain closed, so open alternatives trade off some performance for independence, auditability, and the ability to run without an external dependency.
In megatrends
Finds citing this concept
Overall = breadth + depth + substance + ½·novelty + freshness (a small boost that halves each week)
The empty top of open
The G7's digital ministers agreed on a shared four-rung ladder for what 'open' means in AI — and the top two rungs are ones almost no shipping model can reach.
River AI
Igor Babuschkin, who built xAI's Colossus supercluster, left to sell fine-tuning of open models as the seed of an AI you'll one day own outright — but the only shipped piece runs on his cloud, not yours.
How this connects
Tap a node to open it