What does "mixture-of-experts" mean?

SuperCosmos is a mixture-of-experts (MoE) model. Instead of using every part of the network for every token, it sends each token through a subset of specialised sub-networks ("experts").

• Total parameters: 550B, everything the model has learned.
• Active per token: 55B, the part that actually works on each token.

That's why it can be large and still fast for its size. All three tiers (Flash, Max and Heavy) share the same weights; they differ in how much thinking each step gets.

Good to know#

• Not sure which tier? auto lets the product choose per step. What auto does
• SuperCosmos reads text only; it doesn't take image input. What it can't do
• Insulting or cursing at the assistant earns a warning, then a 24-hour pause on a second strike within 30 days. Abuse policy

Still need help?#

• General support: support@clatterlabs.com or team@clatterlabs.xyz
• Infrastructure questions, or permanent deletion of your data (your whole account or one service): infrastructure@clatterlabs.xyz

We usually reply within the first hour, and always within 24 hours. A full support team is with you.