In my testing, qwen-3.6-27b in full precision is well below sonnet, but above claude haiku in coding tasks. Gemma is not even close to qwen, it’s much, much worse.
When I see a company this big screwing over their partner after 20 years of partnership for some lousy tens of millions of dollars (which is probably peanuts compared to worldwide profits), I immediately think that everything is not as simple as it seems.
In this post, Underwood is obviously a virtuous, hard working victim and the sriracha guys are the villains. I don’t believe that there are good and bad companies and I firmly believe that there is some underlying reason for this situation.
Is it a property though? You are buying a non-exclusive revocable license for specific piece of software, it's not an NFT. You don't own anything, you only get permission to use thing you paid for in a set of specific circumstances. Usually it is written somewhere in the license agreement.
What about multi-gpu setups? Nvidia has nvlink, effectively combining any amount of GPUs into one speedy network, which can benefit both training and inference speed for bigger model sizes. AFAIK, AMD doesn't have anything like that, thus, limiting the usefullness of these chips.
Also, compute stack is too rough right now. Not sure ROCm will be comparable with CUDA in terms of usability in the next couple of years at least.
I really, really want to switch from Vivaldi to Firefox, but I am just too accustomed to the workflow. The interface is too slick and it "just works".
I use vertical tabs with different workspaces, and although I am aware I can do it with extensions or userchrome in Firefox, I've found both variants pretty rough and unpolished -- and I don't have the time and energy to polish stuff that bothers me.
The biggest question for me is why nobody creates browsers based on FF engine and everyone uses chromium as a backbone for the browser. I would love to see something like Vivaldi or nyxt based on FF engine.