That’s funny I literally started the $200/month plan this week because I routinely spend $300+/month on API tokens.
And I was thinking to myself, “How does this make any sense financially for Anthropic to let me have all of this for $200/month?”
And then I kept getting hit with those overloaded api errors so I canceled my plan and went back to API tokens.
I still have no idea what they’re doing over there but I’ll happily pay for access. Just stop dangling that damn $200/month in my face if you’re not going to honor it with reasonable access.
Why wouldn't you assume that it implies the API rates are massively inflated? I don't do anything really but started playing around last week. I put $5 in tokens in to see how long it would last while I play around. It came out at 30min of compute or whatever dir 3hrs of playing around. So my dumb back of the envelope says $10/hr of compute means $90k per year. Sure GPUs are expensive but are they $90k year expensive? Dunno. It's not like the incremental cost of adding GPUs to the inference side are unmoored from hardware incremental costs.
It's what claude code's cli /cost command reported. Like I said I haven't really used this stuff much so a lot of it was reviewing what it was doing and thinking about how to prompt it to do what I wanted it to do. I'm generally pretty skeptical about this stuff but I'm an old fart and decided I should see what the kids are doing rather than yell at clouds and tell people to get off my yard. So far I am able to get it to mostly work acceptably for annoying maintenance tasks (adding docstrings, explain code, rough drafts of tests). Things that I just generally don't have time for. Having it explain my own code from a few years ago to jog my memory about how it works. That sort of thing.
Interestingly, it came up with some definitions of abbreviations and slightly different ways to do things using confidential APIs that have been sort of reverse engineered in public but are not fully understood. Which to me was the most interesting. The things it suggested related to that are very plausible and I was trying to decide if it was clever hallucinations or because it's being used/trained on code bases at that company or by others who have better access. So I spent a lot of time googling to see if it had found some publicly available info to confirm.
Anyway after I hit $5 I subscribed to pro to continue the playing but /cost won't tell you anything useful after switching to pro.
I also wanted to try Gemini clinto compare but I couldn't figure out how to actually give Google my money. Why is Google so impossible to use.
This was with Opus? What sort of tasks were you doing? I found that very large files (2.5MB source file) can really eat up tokens. Other than that, I've never run out with my 100 EUR plan, exclusively using Opus.
And I was thinking to myself, “How does this make any sense financially for Anthropic to let me have all of this for $200/month?”
And then I kept getting hit with those overloaded api errors so I canceled my plan and went back to API tokens.
I still have no idea what they’re doing over there but I’ll happily pay for access. Just stop dangling that damn $200/month in my face if you’re not going to honor it with reasonable access.