Local AI
Qwen3.8-27B Achieves 144 Tokens Per Second on Max M5 Processor
Qwen3.8-27B achieved an inference speed of 144 tokens per second on the Max M5 processor, marking a significant step for local AI execution on mobile hardware.
SSICT News ·
Latest News
Microsoft Build Details High-Performance Local AI Hardware Specifications
Microsoft Build has revealed high-performance hardware specifications that enable developers to run trillion-parameter models locally on Windows devices.
NVIDIA and Microsoft Partner on Agentic AI PC
NVIDIA and Microsoft are reportedly collaborating on a new class of personal computers designed to run artificial intelligence models locally.
Dell Deskside Agentic AI Enables Local AI Workflows from Desk to Data Center
Dell Technologies introduces Dell Deskside Agentic AI, a new solution for deploying agentic AI locally with predictable costs, supporting models from 30 billion to 1 trillion parameters.
Ollama 0.19 Powered by MLX Brings Fastest AI Performance on Apple Silicon
Ollama 0.19 is now powered by Apple’s MLX framework, delivering up to 57% faster prefill speeds and nearly double the decode performance on Apple Silicon. The update also introduces NVFP4 quantization support and smarter caching for coding agents.
Perplexity Personal Computer Brings AI Agents to Mac Mini Files
Perplexity Personal Computer allows AI agents to access files on a Mac mini by coordinating multiple AI models to complete user tasks. The product is currently available via waitlist only.
- Ad