You can't realistically replace a frontier coding model on any local hardware that costs less than a nice house, and even then it's not going to be quite as good.

But if you don't need frontier coding abilities, there are several nice models that you can run on a video card with 24GB to 32GB of VRAM. (So a 5090 or a used 3090.) Try Gemma4 and Qwen3.5 with 4-bit quantization from Unsloth, and look at models in the 20B to 35B range. You can try before you buy if you drop $20 on OpenRouter. I have a setup like this that I built for $2500 last year, before things got expensive, and it's a nice little "home lab."

If you want to go bigger than this, you're looking at an RTX 6000 card, or a Mac Studio with 128GB to 512GB of RAM. These are outside your budget. Or you could look at a Mac Minis, DGX Spark or Strix Halo. These let you bigger models much slower, mostly.

You can buy a roughly $40k gpu (the h100) which will cost $100/mo in electricity on top of that to get about 30-80% the performance of OpenAI or Anthropic frontier models, depending what you're doing.

Over 5 years, that works out to ~$45k vs ~$10k, and during that duration, it's possible better open models will come available making the GPU better, but it's far more likely that the VC-fueled companies advance quicker (since that's been the trend so far).

In other words, the local economics do not work out well at a personal scale at all unless you're _really_ maxing out the GPU at close to 50% literally 24/7, and you're okay accepting worse results.

As long as proprietary models advance as quickly as they are, I think it makes no sense to try and run em locally. You could buy an H100, and suddenly a new model that's too large to run on it could be the state of the art, and suddenly the resale value plummets and it's useless compared to using this new model via APIs or via buying a new $90k GPU with twice the memory or whatever.

For something like OpenClaw you realistically only need rather slow inference, so use SSD offload as described by adrian_b here: https://news.ycombinator.com/item?id=47832249 Though I'm not sure that the support in the main inference frameworks (and even in the GGUF format itself, at least arguably) is up to the task just yet.

You can use several times cheaper models than Claude as well, its not like you need anything big to handle all the uses cases listed above

For something the size of Claude, probably not. But for smaller models, maybe (though they also are much cheaper to buy tokens for)

In my circle you'd get called out for taking on a $350 car payment much less a mobile game.

I would imagine that the list of digital chores of a very busy businessman are a bit more extensive. Even in your list, groceries is something that becomes digital once you're high enough in income.

Another weird cultural quirk of the Dutch that will hopefully go the way of Zwarte Piet one day.

The added cost of having an additional person to provide room and food for way exceeds that €300/month. Especially, when taking into consideration that you might have to extend/renovate the house to lodge another person. Adding an extra bedroom and possibly bathroom is not cheap.

A semi-skilled English-speaking customer service agent in PH makes less than $700 a month to put this into perspective.

Working abroad is a totally reasonable proposition compared to working in the Philippines.

So in reality you’re paying for their food, electricity and heat, letting them rent a room for free, and allowing them the use of the other facilities in your home and on top of that you’re giving them a spending allowance of 300 euro.

The customer they’re picturing in their mind’s eye is obvious. The out of touch part comes in when you look at the size of that market— not big— with how likely that market is to grow drastically— not very— and the amount they’re investing in building the product— all of everything plus a bazillion. With what they’ve invested, if they end up with an institutional market the likes of Microsoft split up among the winners, they fucked up.

The economics of these businesses are based way more on hope and hype than rational analysis and planning.

> models performed much worse for the skills I'm using

Hard to believe unless your are doing something much more complex than the things you listed