Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Cerebras Code

cerebras.ai

Post reply on HN
Print view View original

11–20 of 185 posts

Re: Cerebras Code

#11

Post by social_quotient »

sophia01 wrote: post #3

My understanding is that the coding agents people use can be modified to plug into any LLM provider's API? The difference here seems to be that Cerebras does not appear to have Qwen3-Coder through their API! So now there is a crazy fast (and apparently good too?) model that they only provide if you pay the crazy monthly sub?

Exactly! You can use tools like https://github.com/musistudio/claude-code-router which let you use other LLMs.

The way I would use this $50 Cerebras offering is as a delegate for some high token count items like documentation, lint fixing, and other operations as a way not only to speed up the workflow but to release some back pressure on Anthropic/claude so you don’t hit your limits as quickly… especially with the new weekly throttle coming. This $50 dollar jump seems very reasonable, now for the 1k completions a day, id really want to see and get a feel for how chatty it is.

I suppose thats how it starts but id the model is competent and fast, the speed alone might force you a bit to delegate more to it. (Maybe sub agent tasks)

Re: Cerebras Code

#12

Post by amirhirsch »

thanhhaimai wrote: post #8

> running at speeds of up to 2,000 tokens per second, with a 131k-token context window, no proprietary IDE lock-in, and no weekly limits! I was excited, then I read this: > Send up to 1,000 messages per day—enough for 3–4 hours of uninterrupted vibe coding. I don't mind paying for services I use. But it's hard to take this seriously when the first paragraph claim is contradicting the fine prints.

the distinction is from weekly limits of claude code.

Re: Cerebras Code

#13

Post by sneilan1 »

I'm so excited to see a real competitor to Claude Code! Gemini CLI, while decent, does not have a $200/month pricing model and they charge per API access - Codex is the same. I'm trying to get into the https://cloud.cerebras.ai/ to try the $50/month plan but I can't even get in.

Re: Cerebras Code

#14

Post by alfalfasprout »

2k tokens/second is insane. While I'm very much against vibe coding, such performance essentially means you can get near-github copilot level speed with drastically better quality.

For in-editor use that's game changing.

Re: Cerebras Code

#15

Post by knicholes »

It says it works with your favorite IDE-- How do you (the reader) plan to use this? I use Cursor, but I'm not sure if this replaces my need to pay for Cursor, or if I need to pay for Cursor AND this, and add in the LLM?

Or is VS code pretty good at this point? Or is there something better? These are the only two ways I'd know how to actually consume this with any success.

Re: Cerebras Code

#16

Post by sneilan1 »

thanhhaimai wrote: post #8

> running at speeds of up to 2,000 tokens per second, with a 131k-token context window, no proprietary IDE lock-in, and no weekly limits! I was excited, then I read this: > Send up to 1,000 messages per day—enough for 3–4 hours of uninterrupted vibe coding. I don't mind paying for services I use. But it's hard to take this seriously when the first paragraph claim is contradicting the fine prints.

1,000 messages per day should be plenty as a daily development driver. I use claude code sonnet 4 exclusively and I do not send more than 1,000 messages per day. However, that is my current understanding. I am certainly not pressing enter 1,000 times! Maybe there are more messages being sent under the hood that I do not realize?

Re: Cerebras Code

#17

Post by HardCodedBias »

This has to be a monstrous money loser.

If they can maintain this pricing level, and if Qwen3‑Coder is as good as people say then they will have an enormous hit on their hands. A massive money losing hit, but a hit.

Very interesting!

PS: Did they reduce the context window, it looks like it.

Re: Cerebras Code

#18

Post by sneilan1 »

amirhirsch wrote: post #12
thanhhaimai wrote: post #8

> running at speeds of up to 2,000 tokens per second, with a 131k-token context window, no proprietary IDE lock-in, and no weekly limits! I was excited, then I read this: > Send up to 1,000 messages per day—enough for 3–4 hours of uninterrupted vibe coding. I don't mind paying for services I use. But it's hard to take this seriously when the first paragraph claim is contradicting the fine prints.

the distinction is from weekly limits of claude code.

Claude code weekly limits are hard to distinguish. It's not easy to understand their usage limits. I've found when I run into too much opus usage, I switch to sonnet but I've never ran into a usage limit with sonnet 4 yet.

Re: Cerebras Code

#19

Post by alfalfasprout »

knicholes wrote: post #15

It says it works with your favorite IDE-- How do you (the reader) plan to use this? I use Cursor, but I'm not sure if this replaces my need to pay for Cursor, or if I need to pay for Cursor AND this, and add in the LLM? Or is VS code pretty good at this point? Or is there something better? These are the only two ways I'd know how to actually consume this with any success.

any plugin that allows using an OpenAI compatible endpoint should work fine (eg; RooCode, Cline, etc. for VSCode).

Personally, I use code-companion on neovim.

Maybe not the best solution for vibe coders but for serious engineers using these tools for AI-assisted development, OpenAI API compatibility means total flexibility.

Re: Cerebras Code

#20

Post by thanhhaimai »

sneilan1 wrote: post #16
thanhhaimai wrote: post #8

> running at speeds of up to 2,000 tokens per second, with a 131k-token context window, no proprietary IDE lock-in, and no weekly limits! I was excited, then I read this: > Send up to 1,000 messages per day—enough for 3–4 hours of uninterrupted vibe coding. I don't mind paying for services I use. But it's hard to take this seriously when the first paragraph claim is contradicting the fine prints.

1,000 messages per day should be plenty as a daily development driver. I use claude code sonnet 4 exclusively and I do not send more than 1,000 messages per day. However, that is my current understanding. I am certainly not pressing enter 1,000 times! Maybe there are more messages being sent under the hood that I do not realize?

The issue is not about whether the limit is too high or too low. What turned me back was that they claimed "no weekly limits" as a selling feature, without mentioning that they change it to a "daily limits".

I understand it's a sale tactics. But it seems not forthcoming, and it's hard for me to trust the rest of the claims.

Post reply on HN