Hmm, I think that pricing (1 request) was kept until the government ban, so it looks like the price changed along with today’s comeback. As I’ve mentioned, I get it to an extent, but I’d really appreciate some honesty and transparency.
Claude has Fable usage limits. Does Cursor?
@Brent_Rowe Cursor have it’s own api pool usage limit. And no. Claude draw Fable 5 usage from the included weekly limit up to 50% after which it’s drawn at Api rate. And that’s only until July 7th.
Anthropic’s redeployment announcement for Fable 5 mentions that if a request triggers the new safety classifier, the user will receive a notification and the request will automatically be handled by Opus 4.8 instead.
Could you please clarify how this will work inside Cursor? Specifically:
-
Will there be a visible banner or system message in the chat/composer interface?
-
Or is the switch completely transparent, so we only notice the difference through response quality?
-
Is there any log entry or status indicator we can check to confirm whether we’re actually using Fable 5 or Opus 4.8 for a given response?
We need to know this so we can accurately evaluate the answers and manage our workflow. Looking forward to your clarification. Thanks!
Briefly? You mean three days…
Thanks for the info! Do you think it could do a refactor plan (no code) for a rather simple mobile app within like 200 requests?
I need some Opus prompts left 
I’m not able to select fable in the model picker. Am I missing sth?
hey @m_s3, you might have to go enable it in the settings > models
show me your full list of models if you don’t look for a specific one
Hi Cursor team,
About Fable, Anthropic says:
“”"
Advisor strategy: Claude Fable 5 is available as an advisor model. Faster, lower-cost worker models can call Claude Fable 5 mid-task to check their plan and evaluate their work, leading to improved performance.
“”"
Will that be somehow available in Cursor as well (instead of me manually switching)? Would be a nice option, even with Sonnet as base and Opus as advisor.
Thanks
it is pretty much what sub agents do, you can configure one to be fable and other models can call it to discuss or confirm choices
Yes, I think Cursor’s conversion of requests to tokens for legacy pro plans is a bit of a cheek. So yes, I can understand that Cursor is losing money on legacy users if they use up the 500 requests intensively on the new, expensive models. But this token conversion at $20/500… = $0.04 per request plus a 20% surcharge is really cheeky. I could understand it if they’d said, ‘We’ll keep the old models as they are, and for the new ones we’ll charge, say, 10 requests for a Thinking Request and 5 for a normal request’… that would at least have been a fivefold increase, which, above all, is understandable and predictable. But I find the current arrangement simply completely unreasonable.
waiting for this confirmation as well from the Cursor team
If Fable falls back to Opus 4.8, you will see a message:
Fable 5 hit a safety filter, and the conversation was automatically switched to Claude Opus 4.8. Start a new conversation to continue with Fable 5, or continue this conversation with Claude Opus 4.8.
Hi @Andy_Wong Thanks for the post, to reduce token consumption on Fable 5 it’s a good idea to run it medium rather than high, and I think you’ll find the results are still quite good
OK thanks! Good idea. However, would still be nice to have a built-in feature for this, less “homemade”.
Thanks