41–50 of 223 posts
> Reduce your expectations about speed and performance! Wildly understating this part. Even the best local models (ones you run on beefy 128GB+ RAM machines) get nowhere close to the sheer intelligence of Claude/Gemini/Codex. At worst these models will move you backwards and just increase the amount of work Claude has to do when your limits reset.
> Reduce your expectations about speed and performance! Wildly understating this part. Even the best local models (ones you run on beefy 128GB+ RAM machines) get nowhere close to the sheer intelligence of Claude/Gemini/Codex. At worst these models will move you backwards and just increase the amount of work Claude has to do when your limits reset.
That might incentivize it to perform slightly better from the get go.
Earlier quoted context omitted.
Which while expensive is dirt cheap compared to a comparable NVidia or AMD system.
It's still very expensive compared to using the hosted models which are currently massively subsidised. Have to wonder what the fair market price for these hosted models will be after the free money dries up.
Local models are relatively small, it seems wasteful to try and keep them as generalists. Fine tuning on your specific coding should make for better use of their limited parameter count.
Thanks again for this info & setup guide! I'm excited to play with some local models.
Openrouter can also be used with claude code. https://openrouter.ai/docs/guides/claude-code-integration
we can do much better with a cheap model on openrouter (glm 4.7, kimi, etc.) than anything that I can run on my lowly 3090 :)
But I was really disappointed when I tried to use subagents. In theory I really liked the idea: have Haiku wrangle small specific tasks that are tedious but routine and have Sonnet orchestrate everything. In practice the subagents took so many steps and wrote so much documentation that it became not worth it. Running 2-3 agents blew through the 5 hour quota in 20 minutes of work vs normal work where I might run out of quota 30-45 minutes before it resets. Even after tuning the subagent files to prevent them from writing tests I never asked for and not writing tons of documentation that I didn’t need they still produced way too much content and blew the context window of the main agent repeatedly. If it was a local model I wouldn’t mind experimenting with it more.
> Reduce your expectations about speed and performance! Wildly understating this part. Even the best local models (ones you run on beefy 128GB+ RAM machines) get nowhere close to the sheer intelligence of Claude/Gemini/Codex. At worst these models will move you backwards and just increase the amount of work Claude has to do when your limits reset.
The best open models such as Kimi 2.5 are about as smart today as the big proprietary models were one year ago. That's not "nothing" and is plenty good enough for everyday work.
I'll add on https://unsloth.ai/docs/models/qwen3-coder-next
The full model is supposedly comparable to Sonnet 4.5 But, you can run the 4 bit quant on consumer hardware as long as your RAM + VRAM has room to hold 46GB. 8 bit needs 85.
> Reduce your expectations about speed and performance! Wildly understating this part. Even the best local models (ones you run on beefy 128GB+ RAM machines) get nowhere close to the sheer intelligence of Claude/Gemini/Codex. At worst these models will move you backwards and just increase the amount of work Claude has to do when your limits reset.
Maybe add to the Claude system prompt that it should work efficiently or else its unfinished work will be handed off to to a stupider junior LLM when its limits run out, and it will be forced to deal with the fallout the next day. That might incentivize it to perform slightly better from the get go.