[Alifatisk]

So they added reasoning effort to the Chat and also web search, fantastic move at chat.inceptionlabs.ai, haven't tried it yet. I believe efforts towards faster llms is the right track here. "Intelligence" is not the bottleneck anymore but speed. Talaas and Inceptionlabs are the two interesting labs atm. I'll play around with Mercury 2 and see how it is. The benchmark look promising though, I just wish they made the highest score in bold instead of themselves on every column, but thats a tiny nitpick.

Update, even on the highest reasoning effort, it cannot answer the car wash question correctly.