Cerebras inference is indeed impressive. I was getting 1800 t/s yesterday which is incredible. It is is also incredibly difficult to manage. Utilising all that output is like trying to drink from a fire hose at the moment! Could either of you recommend an agentic set up that I can use in conjunction with Cerebras as a base to build on, for the Metaculus forecasting tournaments?
Between these guys and Groq, it's hard to get excited about them when I can't use them in a production environment. Groq's API is useless with their inference use limits. Oh well, I suppose we'll get there eventually.
Cerebras inference is indeed impressive. I was getting 1800 t/s yesterday which is incredible. It is is also incredibly difficult to manage. Utilising all that output is like trying to drink from a fire hose at the moment!
Could either of you recommend an agentic set up that I can use in conjunction with Cerebras as a base to build on, for the Metaculus forecasting tournaments?
... very impressive inference speed, insightful talk with Andrew. cheers! Groq, Samba, Cerebras (most impressive) .. all going for the speed
They are lighting fast and the voice assistant they made is amazing and free for now at least.
Between these guys and Groq, it's hard to get excited about them when I can't use them in a production environment. Groq's API is useless with their inference use limits. Oh well, I suppose we'll get there eventually.
Lol and you can't actually pay them
They’re for enterprise, right?