Universal Music is launching an AI music platform with ElevenLabs17m ago OpenAI Introduces Data Agent in ChatGPT Work to Analyze Company Data19m ago Robotics Startup Skild AI Hits $100 Million in Revenue Run Rate25m ago Is AI The Silver Bullet For Project Portfolio Management?25m ago An Anthropic researcher just quit, saying OpenAI and Anthropic are 'gambling with our lives'26m ago Cohere Debuts Open-Weight 218B Mixture-of-Experts Machine Translation Model28m ago Bending Spoons agrees to buy Miro for $1.355bn, weeks after Airtable31m ago Winmau And Autodarts Bring Smart Scoring To Your Dumb Dartboard32m ago Typewise orchestrates customer-service AI agents38m ago Zuckerberg tells Sources labs holding back AI models is ‘quite dangerous’47m ago Nintendo's latest sale is 'made possible' by the tariff refunds it's hoarding54m ago Now everyone can put data to work55m ago Instagram now lets you add tagged photos to your profile grid55m ago How To Secure The AI Features Hiding Inside Your Sanctioned SaaS Apps55m ago

Universal Music is launching an AI music platform with ElevenLabs17m ago OpenAI Introduces Data Agent in ChatGPT Work to Analyze Company Data19m ago Robotics Startup Skild AI Hits $100 Million in Revenue Run Rate25m ago Is AI The Silver Bullet For Project Portfolio Management?25m ago An Anthropic researcher just quit, saying OpenAI and Anthropic are 'gambling with our lives'26m ago Cohere Debuts Open-Weight 218B Mixture-of-Experts Machine Translation Model28m ago Bending Spoons agrees to buy Miro for $1.355bn, weeks after Airtable31m ago Winmau And Autodarts Bring Smart Scoring To Your Dumb Dartboard32m ago Typewise orchestrates customer-service AI agents38m ago Zuckerberg tells Sources labs holding back AI models is ‘quite dangerous’47m ago Nintendo's latest sale is 'made possible' by the tariff refunds it's hoarding54m ago Now everyone can put data to work55m ago Instagram now lets you add tagged photos to your profile grid55m ago How To Secure The AI Features Hiding Inside Your Sanctioned SaaS Apps55m ago

Chips & Compute Machine Learning4 min reading time

Xiaomi, AI Cube prototype: Multi-chip local LLM powerhouse

Hacker NewsPublished 16 days ago

Xiaomi unveiled the AI Cube prototype, a compact desktop AI supercomputer powered by three custom chips including the XRING O3, D100, and O100. It delivers 200 TOPS, 1.22 TB/s memory bandwidth, and can run 120-billion-parameter large language models locally without cloud reliance.

More on this story

Xiaomi AI Cube Targets Local LLMs with 1.22 TB/S Near-Memory Bandwidth
14d· Hacker News

#custom-ai-chips #large-language-models #local-inference-hardware #xiaomi #xring-d100 #xring-o100 #xring-o3

More in Chips & Compute

Chips & Compute

DOJ Probes Nvidia’s Tie-Up With Groq on Antitrust Concerns

Covered by 2 sources13h ago

Chips & Compute5 min read

D-Matrix Connects Raptor XPUs to NVIDIA AI Factories via NVLink Fusion

Covered by 2 sources3h ago

Chips & Compute5 min read

Celero raises $275m to bring coherent optics inside AI data centres

The Next Web3h ago