200 points103 comments
IBM says open-weight models now 90% effective for 10% cost
What is IBM CEO says open weight models are now good and cheap on-prem?
IBM's CEO announced that open-weight LLMs can now run on-premise delivering about 90 % of the performance of leading closed models while costing only 10-25 % of the token price. The hardware footprint fits in two or three racks, making it a practical option for enterprises that want to keep AI in-house.
Who it is for
Enterprises looking to avoid cloud fees and keep data local should consider deploying these models for cost-effective, high-performing AI.
How to try it
Visit IBM's AI on-prem page or use IBM Cloud Pak for Data for deployment instructions.