Technology & AI
Google's TurboQuant AI-compression algorithm can reduce LLM memory usage by 6x
Source: Ars Technica
Google Research has unveiled TurboQuant, an AI-compression algorithm that can reduce large language model memory usage by up to 6 times while maintaining performance quality. The technique uses a novel quantization method that converts high-precision model weights into lower-precision formats, making it more feasible to run powerful AI models on consumer hardware and reducing computational costs. This advancement could significantly lower barriers to deploying and accessing large language models.
Read Full Article at Ars TechnicaRelated Articles
OpenAI vs Anthropic Web Search Pricing: Same $10 Sticker, Two Different Bills
OpenAI and Anthropic both list web search at $10 per 1,000 calls, but they bill the retrieved content differently. What one search costs on each API, and the settings that cap it.
EU AI Act Article 73 Serious Incident Reporting Requirements: The Clock Starts When You Find Out
Under the EU AI Act Article 73 serious incident reporting requirements, the deadline does not wait for your investigation to finish. It starts when the provider or deployer becomes aware of the serious incident. Article 73 does give fixed limits of 2, 10 and 15 days, but its main instruction is…
EU AI Act AI Literacy Requirements for Managers: A Softer Rule Meets Live Enforcement
Article 4 still requires providers and deployers of AI systems to support their staff's AI literacy, but since mid-July 2026 no specific level is required. National authorities can enforce it from 2 August 2026.
Gemini API Free Tier Data Privacy: What Google Keeps, and Why Europe Gets Paid Rules for Free
On Google AI Studio and the free Gemini API quota, Google uses your prompts and responses to improve its products, and human reviewers may read them. Paid services don't, Europe gets the paid rules for free, and Search grounding keeps data for 30 days.