Try pointing it to a small codebase, or even ask it to conjure information found online.
replyYou'll see that it quickly gives up. Thing is, they seem to count cached hits as if they were the non-cached tokens.
I wont be subscribing again thats for sure. I am not paying iPhone money for a Xiaomi.
That's what I've been doing. I use crush normally. While the codebase are by no means huge, they're not tiny either.
replyAre you using it in an agentic workflow? Just reading the codebase will consume a lot of cached tokens, but seemingly, z.ai counts these as normal input tokens the way they're rate limiting.
replyI'm not entirely sure what an agentic workflow could mean today but I think so. I use a coding agent (crush), prompt it to brainstorm an implementation with me (or sometimes I know exactly how I want to implement it but ask it to challenge it), correct any wrong assumptions or request the implementation to look differently than suggested if I don't like it. Then finally when I'm positive I've cleared the most important assumptions I ask it to actually write and edit files and run tests and such (this just ends up being a "implement this").
replyWith any model I've tried I've found it to be a huge pain to have it fix things where it made a wrong assumption without the code becoming a mess and burning a lot of tokens. I'm aware that not everyone works like this but I'm still very opinionated on what the end result should look like so I can still work on it without an LLM.