Live data from Hacker News

Claude Code: connect to a local model when your quota runs out

boxc.net

81–90 of 223 posts

Re: Claude Code: connect to a local model when your quota runs out

#81
post #3

Useful tip. From a strategic standpoint of privacy, cost and control, I immediately went for local models, because that allowed to baseline tradeoffs and it also made it easier to understand where vendor lock-in could happen, or not get too narrow in perspective (e.g. llama.cpp/open router depending on local/cloud [1] ). With the explosion of popularity of CLI tools (claude/continue/codex/kiro/etc) it still makes sen…

I think control should be top of the list here. You're talking about building work flows, products and long term practices around something that's inherently non-deterministic. And the probability that any given model you use today is the same as what you use tomorrow is doubly doubtful: 1. The model itself will change as they try to improve the cost-per-test improves. This will necessarily make your expectations non…

You're right. Control is the big one and both privacy and cost are only possible because you have control. It's a similar benefit to the one of Linux distros or open source software.

The rest of your points are why I mentioned AI evals and regressions. I share your sentiment. I've pitched it in the past as "We can’t compare what we can’t measure" and "Can I trust this to run on its own?" and how automation requires intent and understanding your risk profile. None of this is new for anyone who has designed software with sufficient impact in the past, of course.

Since you're interested in combating non-determinism, I wonder if you've reached the same conclusion of reducing the spaces where it can occur and compound making the "LLM" parts as minimal as possible between solid deterministic and well-tested building blocks (e.g. https://alexhans.github.io/posts/series/evals/error-compound... ).

Re: Claude Code: connect to a local model when your quota runs out

#82
post #22

> 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 local models are literally right behind Claude/Gemini/Codex. Check the benchmarks.

That said, Claude Code is designed to work with Anthropic's models. Agents have a buttload of custom work going on in the background to massage specific models to do things well.

Re: Claude Code: connect to a local model when your quota runs out

#83
Some native Claude code options when your quota runs out:

1. Switch to extra usage, which can be increased on the Claude usage page: https://claude.ai/settings/usage

2. Logout and Switch to API tokens (using the ANTHROPIC_API_KEY environment variable) instead of a Claude Pro subscription. Credits can be increased on the Anthropic API console page: https://platform.claude.com/settings/keys

3. Add a second 20$/month account if this happens frequently, before considering a Max account.

4. Not a native option: If you have a ChatGPT Plus or Pro account, Codex is surprisingly just as good and comes with a much higher quota.

Re: Claude Code: connect to a local model when your quota runs out

#84

Or just don’t use Claude Code and use Codex CLI. I have yet to hit a quota with Codex working all day. I hit the Claude limits within an hour or less. This is with my regular $20/month ChatGpT subscription and my $200 a year (company reimbursed) Claude subscription.

You're getting downvoted because people here don't know that the specific agent you pick can pollute your context and waste your tokens. Claude's system prompt is enormous, to say nothing of things like context windows and hidden subagents.

Re: Claude Code: connect to a local model when your quota runs out

#85
post #65

Earlier quoted context omitted.

[flagged]

Are you an NVIDIA fanboy? This is a _remarkably_ aggressive comment!

Not at all. I don't even know why someone would be incentivized by promoting Nvidia outside of holding large amounts of stock. Although, I did stick my neck out suggesting we buy A6000s after the Apple M series didn't work. To 0 people's surprise, the 2xA6000s did work.

Re: Claude Code: connect to a local model when your quota runs out

#86
post #22

> 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.

Yeah this is why I ended up getting Claude subscription in the first place.

I was using GLM on ZAI coding plan (jerry rigged Claude Code for $3/month), but finding myself asking Sonnet to rewrite 90% of the code GLM was giving me. At some point I was like "what the hell am I doing" and just switched.

To clarify, the code I was getting before mostly worked, it was just a lot less pleasant to look at and work with. Might be a matter of taste, but I found it had a big impact on my morale and productivity.

Re: Claude Code: connect to a local model when your quota runs out

#88
post #86
post #22

> 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.

Yeah this is why I ended up getting Claude subscription in the first place. I was using GLM on ZAI coding plan (jerry rigged Claude Code for $3/month), but finding myself asking Sonnet to rewrite 90% of the code GLM was giving me. At some point I was like "what the hell am I doing" and just switched. To clarify, the code I was getting before mostly worked, it was just a lot less pleasant to look at and work with. Mig…

Did you eventually move to a $20/mo Claude plan, $100/mo Claude plan, $200/mo, or API based? if API based, how much are you averaging a month?

Re: Claude Code: connect to a local model when your quota runs out

#89
post #22

> 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.

There is tons of improvements in the near future. Even Claude Code developer said he aimed at delivering a product that was built for future models he betted would improve enough to fulfill his assumptions. Parallel vLLM MoE local LLMs on a Strix Halo 128GB has some life in it yet.

Re: Claude Code: connect to a local model when your quota runs out

#90

Earlier quoted context omitted.

Which takes a $20k thunderbolt cluster of 2 512GB RAM Mac Studio Ultras to run at full quality…

Most benchmarks show very little improvement of "full quality" over a quantized lower-bit model. You can shrink the model to a fraction of its "full" size and get 92-95% same performance, with less VRAM use.

> You can shrink the model to a fraction of its "full" size and get 92-95% same performance, with less VRAM use.

Are there a lot of options how "how far" do you quantize? How much VRAM does it take to get the 92-95% you are speaking of?

Post reply on HN