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

61–70 of 185 posts

Re: Cerebras Code

#61

Post by d4rkp4ttern »

unshavedyak wrote: post #38
sneilan1 wrote: post #33

Earlier quoted context omitted.

Oh so this is not an integrated command line tool like Claude code? I assumed it was something where Cerebras released a decent prompt and command line agent setup. A lot of the value of Claude Code is how polished it is and how much work went into the prompt design.

There is i believe a forked Gemini Code which will work like Claude Code, or so it looks like on Youtube.

Yes it’s called OpenCode and today I was surprised to learn it works with Claude Pro/Max subscriptions:

https://opencode.ai/docs/

Re: Cerebras Code

#62

Post by d3vr »

d3vr wrote: post #59
lxe wrote: post #51

Is this available as cline/roo-code integration? I think it might be on openrouter too.

Cline support added in v3.20.4: https://github.com/cline/cline/releases/tag/v3.20.4 Roo Code support added in v3.25.5: https://github.com/RooCodeInc/Roo-Code/releases/tag/v3.25.5 Cerebras has also been added as a provider for Qwen 3 Coder in OpenRouter: https://openrouter.ai/qwen/qwen3-coder?sort=throughput

BTW you can also go through HuggingFace: https://huggingface.co/Qwen/Qwen3-Coder-480B-A35B-Instruct?i...

Re: Cerebras Code

#64

Post by tjr »

jazzyjackson wrote: post #50
chpatrick wrote: post #39

Earlier quoted context omitted.

It's not like human code doesn't need review.

the usage of vibe coding in my experience is towards those folks who run whatever the AI produced and if it does what they expect without throwing errors they ship it. If it throws errors they plug that back into the chatbot until the code stops throwing errors. The whole point of vibe coding is its working faster than you would on your own. If you're reviewing it carefully and understand how it works, you might as w…

Even if it appears to do what you want, but you don't actually read and understand the code, how do you know it doesn't do something else also? Maybe something you don't want?

Re: Cerebras Code

#65

Post by d4rkp4ttern »

ktsakas wrote: post #7

Does it work with claude-code-router? I was getting API errors this week trying to use qwen3 Cerebras through OpenRouter with Claude code router.

I really wish Qwen3 folks put up an Anthropic-compatible API like the Kimi and GLM/Zai folks cleverly did — this makes their models trivially usable in Claude Code, via this dead-simple setup:

https://github.com/pchalasani/claude-code-tools?tab=readme-o...

Re: Cerebras Code

#66

Post by bdcravens »

HardCodedBias wrote: post #35
kristopolous wrote: post #23

Earlier quoted context omitted.

They are a hardware company. They have a custom chip they are running it on. The $200/month is their "poor person" product for people who can't shell out $500k on one of their rigs. https://www.cerebras.ai/system

I know. These things are unbelievable machines. The people at Cerebras are fearless wrt. taking on difficult hardware challenges. But this will certainly be a money loser. They have likely been waiting for an open source model that somewhat conforms to their hardware's limitations and which gives acceptable recommendations. It looks like they have found it with QWEN. We'll see!

OpenAI lost around $5B last year.

https://www.lesswrong.com/posts/CCQsQnCMWhJcCFY9x/openai-los...

Re: Cerebras Code

#67

Post by dmitrygr »

lvl155 wrote: post #49

Their hardware is incredible. Why aren’t more investors lining up for this in this environment?

Contradictions do not exist. Whenever you think that you are facing a contradiction, check your premises. You will find that one of them is wrong.

Re: Cerebras Code

#68

Post by jdiff »

tjr wrote: post #64
jazzyjackson wrote: post #50

Earlier quoted context omitted.

the usage of vibe coding in my experience is towards those folks who run whatever the AI produced and if it does what they expect without throwing errors they ship it. If it throws errors they plug that back into the chatbot until the code stops throwing errors. The whole point of vibe coding is its working faster than you would on your own. If you're reviewing it carefully and understand how it works, you might as w…

Even if it appears to do what you want, but you don't actually read and understand the code, how do you know it doesn't do something else also? Maybe something you don't want?

Irrelevant in vibe coding. If it walks like a duck and quacks like a duck, you don't go looking for extra heads, eyes, fingers, tongues, or tails. You ship it then throw repl.it under the bus when it blows up.

Re: Cerebras Code

#69

Post by lordofgibbons »

rowanG077 wrote: post #48
clbrmbr wrote: post #22

At $200/month the comparable should be Opus 4 not Sonnet 4.

Not really. With Opus 4 you will burn into the thousand a month with serious usage. I tested it yesterday and 5 hours of use was 60$. If I extrapolate that you will easily hit 1K+.

Are you comparing using opus via API based usage vs opus via the $200/mo plan?

Re: Cerebras Code

#70

Post by Flux159 »

Tried this out with Cline using my own API key (Cerebras is also available as a provider for Qwen3 Coder via via openrouter here: https://openrouter.ai/qwen/qwen3-coder) and realized that without caching, this becomes very expensive very quickly. Specifically, after each new tool call, you're sending the entire previous message history as input tokens - which are priced at $2/1M via the API just like output tokens.

The quality is also not quite what Claude Code gave me, but the speed is definitely way faster. If Cerebras supported caching & reduced token pricing for using the cache I think I would run this more, but right now it's too expensive per agent run.

Post reply on HN