alt Hacker NewsThe truth is all frontier models are closed. This isn't even an open source vs open weights thing. Even if you accepted that open weights are "open" the capital requirements for running your own Kimi 3 model are significant. It's not like gcc, where the binary just works well enough on random hardware.
Replies
There is no requirement to own the hardware. Even if you own your servers and GPUs - you would not own the power plant, real estate, or internet infrastructure to support it.
It's trivial to rent all of the above components in a competitive market under different periods, you may simply rent the token output from someone who put in the effort on this as well.
An open frontier model induces margin compression on training and inference prices across the industry.
This is like claiming that Linux isn't open because you don't have a PC. Models can be open-weight even if you don't have the hardware to run them.
Deepseek V4 Flash works on consumer-grade hardware, and it's extremely good.
I get that argument, but I don't think it is that bad. Open weights that are out of reach for local use are still within reach for groups of people or small to medium companies. Even if you only rent enough cloud GPUs to run the weights for a few hours, you still get to do anything you want with it.
Worst case you can still archive the weights and hope for capable hardware to get affordable. At the same time this archive is the base line for a new frontier lab to start over if required. I see open weights as a pure upside even though I can't run the bigger models in my home lab.
>the capital requirements for running your own Kimi 3 model are significant
You don't need much capital up front, you can just rent servers to run it. It's going to be more expensive than openrouter though because you probably don't have economies of scale if you're the only customer.
If all frontier models got opensourced 5 months after they were made available as closed it would be a great world. And it is a great world because that's exactly what open weight Chinese models are doing. They give you same capacity that was closed, just 5 months after they were released.
I'm hoping that all software will be getting opensource clones on par with original software 5 months after the closed source software release.
Even if I cannot run open weight models on my own hardware, I can choose from a number of different inference providers and choose them based on pricing, reliability, speed, privacy policy, etc. There is a healthy competition.
Also, not a single provider can pull the rug if I'd like to use a particular model.