You can find the information about it here.
Naturally, I’m here to request support for it to be added as a model so that it supports tool calls, thinking, etc. I believe the Max mode can also be switched to this variant for faster responses (because of the increased API rates).
Hey, yes, we are planning to switch to this version very soon.
Looking forward to it then!
Given that Gemini 2.5 Pro is cheaper than Claude 3.5 Sonnet of GPT 4o, are we going to see more reasonable request pricing? Selling Gemini Pro access at the same cost as models 2x the cost seems silly.
Just a quick insight, being an entrepreneur myself.
The pricing model of Cursor is based on the complete costs structure of their company, accounting for everything to build, distribute and operate Cursor. The costs per million tokens of an LLM is just a variable in a complex equation to run a viable business.
We should stand behind the Cursor team with 3 tenets:
Then, we all make our individual choices, because without being privy to their business model, making suggestions around how the company should be run seems inappropriate to me.
Not exactly. It depends on cache situation and context length. When prompt caching is used, sonnet 3.7 can be quite cheaper. If the input is a cache hit, GPT 4o costs the same as Gemini 2.5 pro’s cheaper tier(less than 200k context). And if you surpass 200k tokens of context, you’ll have to pay almost the same as non prompt-cached sonnet 3.7. Definitely more than GPT 4o.
In other words, in longer tasks, it’ll probably end up becoming more expensive than either of these two models.
yep, someone got charged $800 for a days worth of work on reddit