🚨 Cerebras GPT sol
task completed in 1 min 50 second while normally it would take 12 min 20 seconds
upto 8-10 times faster , this will be game changer for all the people who cant wait while ai is working
Thinking Machines released Inkling, an open-weight model, available on a new Inkling Playground in the Tinker console.
> Mixture-of-Experts transformer with 975B total parameters and 41B active parameters.
> It supports a context window of up to 1M tokens.
> Inkling scored 1257 points on Design Arena’s Agentic Web Dev leaderboard (GPT-5.6-Sol got 1260).
1TB VRAM is the safe bet if you want to be on the current frontier.
We found some options that are surprisingly cheap for getting to 1TB. We also expect the memory shortage to get better, and for competition to increase for local AI hardware and that to drive down prices, i.e. $/GB will improve (see M7 Ultra 1.5TB VRAM rumor, interestingly at Apple's first local AI event at Apple Park in June they emphasized the 512GB offering so that's likely to stick around with the M5 Ultra).
Smaller models are also getting significantly better, so even a 128GB MacBook Pro or a 128GB DGX Spark is a great insurance policy.
All this info will be freely available on local dot ai so you can make informed decisions driven by data.
놀란 뿐만 아니라 요즘 대형 배급사 통해서 영화 찍는 감독들 아이맥스 비율 고집하는 거 존나 웃긴 게 전세계에 아이맥스 원본 비율로 볼 수 있는 영화관 30개도 안되는데 미국에만 21개 있어서 역사도 짧은 새끼들이 생각도 짧다는 결론이 나옴 존나 인플루언서 프라이빗 파티 같애