Open-source models didnβt just catch up.
They took over. I fully switched my entire workflow to open-source AI.
No more closed models. No more rate limits. No more black-box nonsense. And Iβm never going back.
#OpenSource#AI#AIModel
Open-source models didnβt just catch up.
They took over. I fully switched my entire workflow to open-source AI.
No more closed models. No more rate limits. No more black-box nonsense. And Iβm never going back.
#OpenSource#AI#AIModels
Claude Fable 5 just dropped.
It beats every other AI model on almost every benchmark.
But the interesting part isn't the scores β it's that the longer and harder the task, the bigger its lead.
AI is moving from answering questions to finishing work.
Computers went from filling a whole room β desktops β laptops β the phone in your pocket.
Same intelligence, smaller footprint. Every single leap.
AI is now running the exact same playbook.
Mixture of Experts: don't fire the whole brain, just the few experts each token needs.
Sparsity: most weights barely contribute, so stop paying to compute them.
TurboQuant: squeeze the KV cache to ~3 bits, no accuracy hit, longer context for free.
The room-sized model era is ending. Intelligence in your pocket is just getting started.
π https://t.co/Tvm0qng8OI
Mixture of Recursions is amazing. Google laid a new and optimised path for the next set of frontier models. Here is the link
https://t.co/vRB2sCoeT4
#GoogleAI#Transformers#artificalintelligence
@supabase We want to build a wellbeing app not with typical nosql but with SQL with faster iterations. We finally chose it & loved it.
Thanks @supabase .