I’ve only seen their propaganda, guys do you know his context?
200k or 1M?
They just added it its also 120k tokens, but we will have to see after it stabilizes and they test and update everything.
Found it hahaha, looks like the context hasn’t changed
Yes, it’s currently not working and needs to be fixed
They are on it, im in the call.
1 Like
Thinking model works already.
I am Claude, powered by claude-4-sonnet-thinking.
120k for premium mode.
200k for MAX MODE
Ok, Im using max mode and its upsetting. It is making a lot of calls and then fails.
So I am paying for usage and it does not even finish. This is not acceptable. I should not be charged if it does not complete the work
It makes no sense for me to get charged for it to think and then crash and burn.
With any new model it may be necessary to test prompts again and see how it is different. Also there is a heavy demand now with those models, likely thats why Anthropic cant add more servers for the new models, specially opus which is much more resource intensive.
That is understandable, but my point is that if it times out or fails to deliver, we should not get charged.
I should not have to pay for it to start something that it is unable to finish because their servers can’t handle the load.
5 Likes
Can verify it is actually (on the backend, the REAL context window) 55k for non max, and 98k for max. You can verify what I am saying if you go to [r,e,d,d,I,t].