I have gone down this route but I'm afraid that 16K of context is nothing for agentic coding tasks. Even if you use a smarter model directly from an inference provider to plan all the work and then use your own 16GB GPU to execute the plan, managing those 16k tokens becomes unusable at some point even with a minimalistic harness like pi. For other agentic tasks it's nice, and more than just the budget, it's the data sovereignty that you gain (imagine analysing tax forms, contracts, other legal documents, etc.) in my opinion.

> llama.cpp serves it over localhost at speeds that stop being a complaint after a few minutes of use.

"That stop being a complaint"? That's a strange construction.

I wonder if qwen 4 will make these smaller cards more viable by allowing the ngram storage to be hosted on CPU RAM.

Thanks for writing and sharing. Since I also have a 4070 Ti Super, I'm always interested in hearing people's experiences with local models that fit 'middle-ground' hardware. I kind of feel left out because most posts I see seem to either be about clever ways of making older, lower-end cards usable, leveraging mega-CPU RAM (128GB) or how aweseome high-end GPUs like 4090/5090 can be.

It's two different things to be honest.

Local model can be a local assistant, take care of your todos, calendar, private stuff in general. Things you shouldn't share with OpenAI or other Big Tech. Big Tech models for anything else.

I use the free versions on mobile, and openrouter on desktop. It’s way cheaper than paying a subscription, and I can always use the latest models.

OP could be doing a lot better with a MoE model (uses some RAM as well).

I'm doing Qwen 3.6 35BA3B-NVFP4 (a larger ~20GB model) in 8GB VRAM and 20 GB RAM (I also get more than 16k context, and more t/s).

While I'm sure the goalposts will move, if I could get Opus 4.8+ intelligence with 100tps+ and 1M context window I think I'd be happy to switch, mostly, to local inference.

I cannot wait until we get "AI Appliance"-type things we can just plug in and use (aka a server). I know there are a few examples of this but few are plug-and-play currently. I'm thinking more "Here is your hardware Opus 5.5, it costs $X,XXX, it can do 100tps and X parallel streams with a 1M context window".

Maybe I'll change my tune when that's available due to the frontier being, potentially, still 6mo+ ahead with faster inference and larger context windows but I do pine for local inference for at least the "Execute the plan a bigger/better/smarter model wrote for you and the same model will check your work". Currently I'm trialling DS Flash 4.1 as my "worker" model and seeing positive results with it being wrapped by Opus 5.5 High.

20$/month vs. 1200$ gpu. That's a lot of months you can pay the subscription

Ok. What can I do with an M4 MBP, with 48G memory?

If you were only paying $20 for LLMs previously then you were getting very low usage / little done.

Running QWen costs so much locally, that a $20 subscription costs the same with better quality, significant better but less privacy.

This is a great article! Thank you for sharing. I also recommend opencode that has built in features for local model hosting.

"I spent $1100 on a graphics card so I can save $20/mo running a mid/low intelligence model at 33tk/s with 16k context"

I strongly feel that there’s a certain ideological bend towards self hosting and that it clouds judgement.

Dude I can run Astra on my 20 dollar subscription. The leverage with that vs whatever Chinese model you can run in one GPU is not even comparable.

I almost always reach for the most performant model for any of my tasks because my return almost always pays off.