33
Tell HN: Avoid Cerebras if you are a founder
I was an Enterprise customer on their platform. Cerebras began terminating production models and replacing them every few months. This time they notified us they are terminating Llama 3.3 70B, which is the model my plan was subscribed to.
Instead of offering us an alternative plan they told us all others are sold out and have kicked us off their platform.
Their Discord support group has multiple Enterprise customers all getting the same treatment. Their own support staff is telling them to migrate to Groq.
You can’t build a stable business with a company that randomly terminates models this frequently. Who does not respect their customers (especially Enterprise customers, not even small utilizers), offering zero contingency plans if they decide to axe your model.
Because of their architecture, they can only host a finite number of models. Turnover is fast, so if your model gets marked for deprecation, you will get forced off the platform.
Cerebras is cool for personal Hobby projects, but is in absolutely no position to be selling "Enterprise" accounts to businesses.
Hardware company LARPing as infrastructure provider. Their wafer-scale chips can't multi-tenant like GPUs, so "enterprise" means "first in line for deprecation" apparently. Cool tech, zero operational maturity. Stick to providers who understand that "enterprise" means contracts, not vibes.
I noticed this behavior when the very popular Qwen models suddenly were migrated to GLM
I don't know how many times this has happened in the last 12 months but I'd guess like 2-3 times?
Didn’t Cerebras announce a deal with OpenAI recently? Perhaps they need the resources for that contract…
Surely this isn’t how it was supposed to happen.
Is their runway money running out?
The AI market is extraordinarily skewed towards large players. It's hard for a business to support 100s of 6 figure USD sized accounts next to 1-2 9 figure accounts.
This makes me suspect that OP would have similar problems with other inference providers.
Potentially.
Cerebras in Discussions to Raise Funds at $22 Billion Valuation - https://www.bloomberg.com/news/articles/2026-01-13/cerebras-... | https://archive.today/bz9pI - January 13th, 2026
Cloud service providers deprecate services all the time. To be honest, why are you still on llama 3.3 70B in January 2026? Are you in the Strava AI team?
Do cloud providers kick you off the platform and terminate your account when they deprecate models too?
Did they really do that?
Yes, every single Enterprise customer that was unfortunate enough to be subscribed to LLama 3.3 has had their plans terminated. Everyone would have gladly migrated to OSS or something, but they never gave us the option.
my guess is they are aiming for the high rollers and not shrimp like you even though you were "Enterprise". I worked for a company that wanted to get rid of the lower tier customers to focus the money makers.
I had the second highest tier. But yes, that makes sense. Regardless, I just don't see them as a stable business partner. I know the same cycle will continue with the next round of deprecation. Perhaps the highest tier enterprise users will form a different opinion.
[dead]
Sincere condolences to the Strava devs: if the migration headache gets to be too much, Fireworks AI is still hosting Llama 3.3 70B (and 40+ others). Even if your enterprise budget is more "scrappy startup" than "big tech" right now, our GTM team is open to talking. Or just DIY it—the platform is built for self-serve. If you need some credits to skip the procurement hurdles, email marketing [at] fireworks.ai and we'll hook you up.