alt Hacker News> They don't want you to have access to frontier models. And you will not have it. See Mythos as Exibit A.
"They" fully well know that they current frontier model are maybe 6 month ahead of what people will have access to without their control. See Deepseek as Exibit B
The reason you can't run these locally are more with the fact that those mythos sized models require extreme amount of memory and processing power to run at acceptable speeds. And neither you, nor I can afford to pay for those resources to run those models locally. A big reason is that "running locally" means running on your own hardware. And for almost everyone this means "running on hardware that will spent a big portion of its time just sleeping". Because data center and providers have higher utilization rates, they can easily outpace you. That and the fact that when they place an order it's usually for hundreds of thousands of units.
Replies
Although, I wonder how many orders of magnitude in terms of affordability the utilization rate actually gets them. Realistically if you use a self-hosted LLM for your job, you might be using it, what, a solid 6 hours per day? Assuming you can keep it actually fed, while working (so, some agentic thing might be necessary, I guess it will need to be more than VSCode autocomplete and responding to individual prompts). Anyway, that starts you out at 1/4’th the utilization, a 4X price increase might be worth paying for privacy and stability (no sudden change in model behavior, no price changes, no days when the system is over-utilized for reasons outside your control).
Rather I think it is just hard for local LLMs to compete in this early stage when the cloud providers are allowed by investors to be unprofitable.
Isn't Mythos that screw up where Anthropic failed to ship something that was no better than the product OpenAI launched a few weeks later?
And, assuming the allegations are true, don't things like Deepseek and Qwen offer existence proofs that frontier models are (and will forever be) trivially distilled down to run domain-specific tasks on boxes that cost a few months of Claude Max subscription?
I am convinced the (mainly chinese) open weights models are the only reason OpenAI and Anthropic release at the pace they do. Without them being on their heels, we would have seen a stagnant duopoly in terms of public releases.
That is why the huge lobby machine is grinding away to make those models illegal.