GptZone

@gptzone_net

Las Mejores Noticias, Prompts, Apps e Hilos sobre Inteligencia Artificial! 📩 DM Abierto Suscríbete a Nuestra Newsletter Diaria 📌https://news.gptzone.net

following
followers

Rock Matrix open-sources EdgeRock-4B model with efficient KV cache memory

Rock Matrix released EdgeRock-4B, a ~4B parameter text-and-image model reportedly using a fixed-size KV cache to limit RAM growth.

7 Sources, 9d ago, updated 9d ago

Perplexity updates voice mode to work across multiple Computer sessions

Perplexity says Pro and Max subscribers can start tasks and manage Projects in one voice session on Mac and web.

10 Sources, 12d ago, first seen 12d ago

Perplexity unveils Decider for Mac to run its 4B decision model locally

Perplexity says the app answers questions about text and images through a local API without a usage bill.

13 Sources, 16d ago, updated 16d ago

Canva introduces Craft, its first frontier image model built for design

Canva claims it ranks second among AI labs for image generation on DesignArena’s global leaderboard.

19 Sources, 17d ago, updated 17d ago

Artificial Analysis announces AA-PersonalityBench to compare how AI models write

Artificial Analysis uses 900 questions to assess 30 judged traits and 21 separately measured writing characteristics.

31 Sources, 18d ago, updated 18d ago

Featured in

Rock Matrix open-sources EdgeRock-4B model with efficient KV cache memory

Perplexity updates voice mode to work across multiple Computer sessions

Perplexity unveils Decider for Mac to run its 4B decision model locally

Canva introduces Craft, its first frontier image model built for design

Artificial Analysis announces AA-PersonalityBench to compare how AI models write

Featured in

Rock Matrix open-sources EdgeRock-4B model with efficient KV cache memory

Perplexity updates voice mode to work across multiple Computer sessions

Perplexity unveils Decider for Mac to run its 4B decision model locally

Canva introduces Craft, its first frontier image model built for design

Artificial Analysis announces AA-PersonalityBench to compare how AI models write