LIVE

Edge0 Releases 145B On-Device AI Model to Rival Cloud Capabilities

🧠 AI Analysis 2 min read 👁 11

Developers have launched Edge0-145B, an open-source on-device AI model with 145 billion parameters designed to run locally on personal computers.

Edge0 Releases 145B On-Device AI Model to Rival Cloud Capabilities🧠AI Analysis

Photo: Hacker News Front Page

Key points

  • Edge0-145B is introduced as an open-source, on-device AI model with 145 billion parameters.
  • The architecture uses shared matrices among sparse experts so only ~10B parameters are active per inference step.
  • A tiered storage pipeline coordinates GPU memory, system RAM, and SSD access to manage local workloads.

A new open-source initiative has introduced Edge0-145B alongside a smaller 120B variant and a unified inference framework, aiming to bring advanced coding, reasoning, and agent capabilities directly to local hardware. Historically, running large artificial intelligence models on personal devices has required balancing hardware limits against processing power, often forcing users to rely on cloud providers for demanding tasks.

To overcome memory and bandwidth constraints, the creators engineered the model specifically for local execution from the start. Edge0-145B utilizes a shared matrix architecture among sparse experts, meaning that out of its total 145 billion parameters, only about 10 billion participate during each individual inference step. The deployment strategy also pipelines data access across GPU memory, system memory, and solid-state storage to maintain performance without overwhelming local hardware resources.

The release includes the complete open-source stack, allowing developers to reproduce findings, evaluate and fine-tune weights, or integrate the models into personal applications. Previous iterations, such as the Edge0-35B model, demonstrated efficient memory profiles on mobile hardware, while the new 145-billion-parameter system reports high operational speeds and a peak memory usage of roughly 28GB on laptop testing environments.

What to watch next: Monitor developer adoption of the Edge0 framework and independent hardware benchmarks on consumer laptops.

Source Hacker News Front Page — edge0.ai/blog/introducing-edge0-145b-frontier-ai-capabilities-on-your-o… ↗

Contact the newsroom

Got a tip, a proposal or a partnership question? Write to us and we will reply shortly.