Sublime

An inspiration engine for ideas

AllPeopleCollectionsArticlesAudioBooksFilesHighlightsImagesLinksNotesTextTweetsVideosSocial

REAL-TIME object detection WITHOUT TRAINING YOLO-World is a new SOTA open-vocabulary object detector that outperforms previous models in terms of both accuracy and speed. 35.4 AP with 52.0 FPS on V100. ↓ read more https://t.co/SoqFyEk41V

SkalskiP x.com

TogetherAI just released StripedHyena 7B. At first look ~Mistral level model with significant memory and speed benefits: https://t.co/ihZug2NuRS If you are interested in how it works, I surveyed this literature earlier this year in an MLSys keynote: https://t.co/xTW79oYQ6P

Sasha Rush x.com

Clarity AI is 100% open-source and can be self-hosted while Magnific is closed: https://t.co/gEwPd6ltJw Clarity AI is free to use on Replicate, while Magnific is paid only: https://t.co/uuefmgK724 Clarity AI has an API while Magnific does not, making it unable to... See more

@levelsio x.com

Introducing Real-time Transcription with Nvidia Parakeet - Same top accuracy as file transcription - Best-in-market 160 ms lips-to-screen latency - 744x more cost-efficient compared to cloud APIs - Available in Argmax Pro SDK starting today! Link in comments

argmax x.com

The best open-source AI model just dropped a detailed report on how it was trained, a rare resource for students given no frontier lab is publishing! Kimi K2's estimated total cost of training is ~$20-30M, roughly in line with pricing: $0.6/M in $2.5/M out tokens. 10 highlights:

Deedy

x.com

Run Kimi K2, the world's most powerful non-reasoning model locally on your computer. Unsloth team shrank the full 1.1TB model to just 245GB (-80% size reduction) making it 1.8-bit. 100% Opensource and free. https://t.co/OehjvWCMl4

Shubham Saboo

x.com

my detailed personal benchmarks ran overnight. - Scout is best at summarization and function calling. exactly what you want from a cheap long ctx model. this is going to be a workhorse in coding flows and RAG applications. the single shot ICL recall is very very good. - Maverick was built for replacing developers and... See more

xjdr x.com

Thumbnail of www-x-com-hubermanlab-status-1828842248378953807-3de90eafa241404c

This is really cool. (If you are interested in dogs, animal behavior, dog genetics etc.) As discussed on @joeroganhq with @joerogan https://t.co/uEyMKVEERF

Andrew D. Huberman, Ph.D.

x.com

New 7B model now beats LLaMA 2 70B completely Beautiful work! TL;DR: 1. Started from Mistral 7B. 2. SFT on filtered UltraChat dataset: 200K samples [1] 3. Trained from feedback (DPO) on similar version of the dataset [2] for helpfulness only: 64K... See more

Yam Peleg x.com