Open sourceNov 12, 2026, 6:06 PM EST
Edge0 open-sources 145B on-device model for laptops
The company reports the 145B model needs about 29GB peak memory, decodes at over 80 tokens per second, and supports context of 128K to 256K tokens. Model weights, the framework and an Edge0 App are available now for developers to reproduce results, fine-tune the models, or build them into products. Edge0 says local inference carries no cloud token or API fees.
Why it matters
Developers who want coding and agent-style work done without sending data to a model provider now have a locally runnable option, at the cost of a compatible machine with substantial memory and roughly 120W typical power draw.
Sources
- PrimaryHacker News
Written by AIToolMaps from the source above and checked against it sentence by sentence.