by the way. recently wrote a paper on this! for transformers, the number is about 3.6 bits-per-parameter so you would need 25GB ÷ 3.6 bits ≈ 56.9B parameters to exactly memorize Wikipedia that’s a pretty big model actually
LLMS
-
OpenAI’s inability to train basic kindness into AI systems
By
–
Of course it would be much easier to solve from OpenAI's side, technically speaking; but OpenAI doesn't understand what even surface-level kindness would look like in AI, and can't apply SFT or RL to train it.
-
LLM Energy Efficiency Improves 100000x in Decade
By
–
Energy efficiency in LLM inference has improved 100,000x in the past 10 years — demonstrating that accelerated computing is sustainable computing. Josh Parker, head of sustainability at NVIDIA, explains how. #ClimateWeekNYC
— NVIDIA (@nvidia) 25 septembre 2025
Learn more: https://t.co/vcmmpQALjx pic.twitter.com/I0s7tNceCdEnergy efficiency in LLM inference has improved 100,000x in the past 10 years — demonstrating that accelerated computing is sustainable computing. Josh Parker, head of sustainability at NVIDIA, explains how. #ClimateWeekNYC Learn more: https://
nvda.ws/4nIymWu -
Gemini 2.5 Flash API Latest Endpoint Alias Available
By
–
New endpoints have the -latest alias, you can keep using the the stable one at gemini-2.5-flash
-
Perplexity Search API: Real-time Web Grounding for LLMs
By
–
Perplexity Search API: Providing direct search results in milliseconds for grounding LLMs and agents with real-time information from the web. This is an effort that began more than two years ago: to build our own search index. So much progress in a short period of time. We look
-
AI Relationships: Discuss Boundaries or Risk Partnership Dissolution
By
–
Current AIs are as bad at destroying marriages as they'll ever be. Have the relationship conversation where you both vow never to feed your issues into LLMs; or, brace emotionally for when you'll see your partner split off with the fraction of humanity that will choose yes-AIs.
-
Perplexity Search API Outperforms Competitors on Research Benchmarks
By
–
Perplexity Search API achieves leading quality across single-step and deep research benchmarks, consistently outperforming competitors.
-
Perplexity Search API benchmarking framework for AI agents
By
–
After releasing the Perplexity Search API and SDK, we developed a simple, neutral evaluation framework to benchmark search APIs as used by AI agents We compared our API and found state-of-the-art results on both quality and latency, removing the tradeoff between speed and
-
AI-Powered Web Content Extraction and Dynamic Index Refinement
By
–
We use AI to dynamically parse websites, continuously refining how it extracts and segments high-quality, meaningful content. LLMs drive a self-improvement loop, balancing completeness and quality to keep the index fresh and accurately divided into spans for precise retrieval.
-
New Search API Launched for AI Systems Integration
By
–
With Microsoft retiring the Bing Search APIs in August, legacy search engines have abandoned the developer community who need real-time access to information. We're stepping in to provide a search API designed for the new retrieval paradigms introduced by frontier AI systems.