Local LLMs are the way to go on AI, they can’t suddenly change the model up on you. And also, I guess Anthropic just wants their models to be more useless ahead of going public? Because I presume people that want to use AI, who don’t care where they send their data to, but don’t want to nagged about how they can’t do benign programming and analysis tasks, will just choose another model provider.
They are the way to go, but hardware pricing makes it cost-prohibitive to match frontier capabilities, which makes decomp and recomp difficult. Or rather, they require more skill and take longer, introducing some barriers against the current floodgates being flung open.
We all know the eye-watering prices GPUs and RAM cost now, to say nothing of Unified Memory devices like the Strix. Most people just don’t have a spare $10K to throw at an inference box that can run (say) a 300B model at interactive (not batch) speeds.
Coming at this from another angle tho, I note that according to Steam stats, the most popular consumer GPUs are in the 8GB and 16GB range…which is smack bang in the VRAM range of what a good MoE (say, Qwen 3.6 35B-A3B) runs on (for 8GB) or arguably the best local dense coder (Qwen 3.8 27B) can live at. Which leaves us with a capable assistant and a capable coder.
I think there’s probably room for LLM improvement in the 8–16GB range too…and given that most consumer PCs and laptops live there, I’m sanguine the open-weight devs might make it happen.
But I also suspect the gatekeeping of high-end capacities is intentional.
We’re probably not going to have the ability to vibe code Mario into Elden Ring again…not without significantly more skill. Dunno.
Uhm yeah good luck with running anything locally after hardware prices double again and again. Big tech needs to be stopped. There is no wiggling out of this just by being a smart customer.
The price did grow several times in size, but even in the 4 to 5 digit price range, you or smaller companies can have a pretty competitive performing open source model as the frontier models from earlier this year and they are keeping pace.
Obviously I’d like the speculation bubble to end and the DRAM and GPU cartels broken up, as it’s late-stage capitalism where the existing regulations being unenforced on big tech because the regulators are in on it that’s the problem, not so much the AI tech itself. (Don’t get me wrong, the tech has bad sides to it too like stealing artwork, data centre environmental impact and whatnot, but just speaking to the cause of the component greedflation).
Yes it’s disgustingly expensive. I spent a whole lot of money on an old used GPU and it felt bad, but it didn’t feel dirty like paying for Claude would. And then a couple of weeks later the used prices for this GPU had risen by about 15%.
(I only use Claude at work when someone else is paying. And they’re about to make me redundant because Claude is cheaper than me.)
With an iGPU results are going to be very slow. Most promising things I have heard of so far is Mixture of Experts type models, if you have an 8GB card you might be able to run Qwen 3.8-35B-A3B-MoE which I think is the most capable for that hardware.
There are websites that can suggest a few for your specific hardware, I assume some are vibecoded and are only occasionally updated to the latest models, so take it as you will. See this Lemming’s comment. And !localLLaMa@sh.itjust.works is where I think most of the discussion around new development with Local LLMs happens here on Lemmy.
Another good option is Qwen3.5-9B. I’d try a Q4 or Q5.
Whichever model you choose, I’d also suggest a few MCP servers to give it access to reliable information rather than rely entirely on its internal knowledge.
Microsoft Learn MCP: Official PowerShell, Windows and network documentation. Will help with your scripting stuff.
I’d start with Microsoft Learn and Exa, then add the others as needed. You’ll need an MCP-capable client (which pretty much all of them are now - eg: llama.cpp web ui, OpenWebUI, Light Local Chat etc), and I’d keep file access read-only initially.
I’d personally use McProxy instead of adding MCPs singularly, but I fear I’m already over loading you with information.
EDIT: The PolyStrata recommendation assumes you have a discrete GPU with sufficient VRAM (unclear if you do?). If not, disregard Ornith and try Tiny Ling 3 Q6 or Qwen3.5-9B with ik_llama.cpp instead.
Either way, both smaller models should be capable of the tasks you’ve described, particularly with access to the right documentation. Just verify any network or firewall changes before applying them.
Local LLMs are the way to go on AI, they can’t suddenly change the model up on you. And also, I guess Anthropic just wants their models to be more useless ahead of going public? Because I presume people that want to use AI, who don’t care where they send their data to, but don’t want to nagged about how they can’t do benign programming and analysis tasks, will just choose another model provider.
They are the way to go, but hardware pricing makes it cost-prohibitive to match frontier capabilities, which makes decomp and recomp difficult. Or rather, they require more skill and take longer, introducing some barriers against the current floodgates being flung open.
We all know the eye-watering prices GPUs and RAM cost now, to say nothing of Unified Memory devices like the Strix. Most people just don’t have a spare $10K to throw at an inference box that can run (say) a 300B model at interactive (not batch) speeds.
Coming at this from another angle tho, I note that according to Steam stats, the most popular consumer GPUs are in the 8GB and 16GB range…which is smack bang in the VRAM range of what a good MoE (say, Qwen 3.6 35B-A3B) runs on (for 8GB) or arguably the best local dense coder (Qwen 3.8 27B) can live at. Which leaves us with a capable assistant and a capable coder.
I think there’s probably room for LLM improvement in the 8–16GB range too…and given that most consumer PCs and laptops live there, I’m sanguine the open-weight devs might make it happen.
But I also suspect the gatekeeping of high-end capacities is intentional.
We’re probably not going to have the ability to vibe code Mario into Elden Ring again…not without significantly more skill. Dunno.
Uhm yeah good luck with running anything locally after hardware prices double again and again. Big tech needs to be stopped. There is no wiggling out of this just by being a smart customer.
Thing is, I already am…
The price did grow several times in size, but even in the 4 to 5 digit price range, you or smaller companies can have a pretty competitive performing open source model as the frontier models from earlier this year and they are keeping pace.
Obviously I’d like the speculation bubble to end and the DRAM and GPU cartels broken up, as it’s late-stage capitalism where the existing regulations being unenforced on big tech because the regulators are in on it that’s the problem, not so much the AI tech itself. (Don’t get me wrong, the tech has bad sides to it too like stealing artwork, data centre environmental impact and whatnot, but just speaking to the cause of the component greedflation).
Yes it’s disgustingly expensive. I spent a whole lot of money on an old used GPU and it felt bad, but it didn’t feel dirty like paying for Claude would. And then a couple of weeks later the used prices for this GPU had risen by about 15%.
(I only use Claude at work when someone else is paying. And they’re about to make me redundant because Claude is cheaper than me.)
Is there a local LLM I can run on an 8-10th gen i5 with 16GB RAM?
With an iGPU results are going to be very slow. Most promising things I have heard of so far is Mixture of Experts type models, if you have an 8GB card you might be able to run Qwen 3.8-35B-A3B-MoE which I think is the most capable for that hardware.
There are websites that can suggest a few for your specific hardware, I assume some are vibecoded and are only occasionally updated to the latest models, so take it as you will. See this Lemming’s comment. And !localLLaMa@sh.itjust.works is where I think most of the discussion around new development with Local LLMs happens here on Lemmy.
links
Yes, quite a few, but it depends on what you’re trying to do with them. Any ideas?
Assist with writing letters, home assistant code, light powrshell scripting.
Edit: helping segment my network
That should be well within reach of a local LLM.
If you’ve got a discrete GPU with sufficient VRAM, something like Ornith running on PolyStrata could be worth a shot. Try a Q3 size.
https://huggingface.co/bartowski/Ornith-1.5-35B-A3B-GGUF
https://github.com/VecSzn/PolyStrata
Hell, Tiny Ling 3 might be able to do it. Try a Q6.
https://huggingface.co/bartowski/Ling-3.0-tiny-GGUF
Another good option is Qwen3.5-9B. I’d try a Q4 or Q5.
Whichever model you choose, I’d also suggest a few MCP servers to give it access to reliable information rather than rely entirely on its internal knowledge.
https://learn.microsoft.com/en-us/training/support/mcp
https://github.com/exa-labs/exa-mcp-server
https://github.com/modelcontextprotocol/servers/tree/main/src/filesystem
If you’re actually running Home Assistant, its official MCP server is worth a look too.
https://www.home-assistant.io/integrations/mcp_server/
I’d start with Microsoft Learn and Exa, then add the others as needed. You’ll need an MCP-capable client (which pretty much all of them are now - eg: llama.cpp web ui, OpenWebUI, Light Local Chat etc), and I’d keep file access read-only initially.
I’d personally use McProxy instead of adding MCPs singularly, but I fear I’m already over loading you with information.
EDIT: The PolyStrata recommendation assumes you have a discrete GPU with sufficient VRAM (unclear if you do?). If not, disregard Ornith and try Tiny Ling 3 Q6 or Qwen3.5-9B with ik_llama.cpp instead.
https://github.com/ikawrakow/ik_llama.cpp
Either way, both smaller models should be capable of the tasks you’ve described, particularly with access to the right documentation. Just verify any network or firewall changes before applying them.
They suck far too much at the moment.