Local models are coming to your laptop soon! ๐
We're excited to partner with @Microsoft to enable millions of developers run local models on Windows!
hot take: capability has stopped being the interesting axis for picking an LLM.
it's table stakes now. three frontier models can all clear the bar your task needs.
the real question moved one layer up: which one tells you when it's wrong, and which one burns your latency budget
so when people ask which model i'd pick for a production agent in 2026, it's almost never "the smartest one."
it's the one that says "i'm stuck" instead of faking completion. the one that answers in 200 tokens instead of 4,000. that axis isn't on the leaderboard you're reading.
what started as a side project is now ranked #2 on the BRIGHT benchmark edging out Nvidia nemo
trained a 4B embedding model on a single 3060 ti using @UnslothAI
Top 3 submissions:
INF-X-Retriever: 63.4
RakanEmbed: 52.4 (uses INF-X rewritten queries)
NVIDIA NeMo: 50.9
Disheartened and enraged to hear the misogynistic comments made by Mr.Mansoor Ali Khan, given that we all worked in the same team. Respect for women, fellow artists and professionals should be a non-negotiable in any industry and I absolutely condemn this behaviour.
A recent video has come to my notice where Mr.Mansoor Ali Khan has spoken about me in a vile and disgusting manner.I strongly condemn this and find it sexist,disrespectful,misogynistic,repulsive and in bad taste.He can keep wishing but I am grateful never to have shared screen space with someone as pathetic as him and I will make sure it never happens for the rest of my film career as well. People like him bring a bad name to mankind.