Urgent.News

What's breaking now, across hundreds of outlets.

AI

Edge0-145B: Frontier AI Capabilities on Your Own Computer

Article URL: https://edge0.ai/blog/introducing-edge0-145b-frontier-ai-capabilities-on-your-own-computer Comments URL: https://news.ycombinator.com/item?id=50475713 Points: 333 # Comments: 101

Frontier AI capabilities are now accessible on personal computers with the open-source release of Edge0-145B and Edge0-120B models. These powerful models bring coding, reasoning, and agent capabilities comparable to large cloud models directly to users' devices.

Edge0-145B, with 145 billion parameters, offers a 4x increase in parameters compared to its predecessor, Edge0-35B, which delivered 37+ tokens per second on an iPhone 18 Pro Max using approximately 2.5GB of memory. Edge0-145B achieves 80+ tokens per second and requires 28GB of peak memory on a laptop.

The release addresses memory, bandwidth, power, and latency constraints by releasing the full stack, including model weights, inference framework, and deployment tools. Developers can reproduce results, fine-tune the models, and integrate them into their own products.

The design of Edge0-145B incorporates a unified inference framework and deployment tools that optimize memory usage, reduce data movement, and balance GPU, system, and SSD memory access. By sharing matrices across sparse experts, the architecture significantly reduces the weight storage and movement required for computation.

Benchmark results show Edge0-145B performing comparably to specific frontier cloud models in coding, reasoning, knowledge, and instruction following tasks. Its capabilities extend to local workflows that combine planning, tool use, execution, debugging, and visual checks.

Rewritten by urgent.news; may contain errors — check the original.

Source: edge0.ai

This story

This is one outlet's version. Read the fullest account.

Every outlet that reported this, and when →

Read the original at edge0.ai →

Report a problem with this story

More in AI

More from Thursday 12 November →