Original post
Marco Mascorro#598
Marco Mascorro#598Cursor integrates Claude Fable 5 to hit a record 72.9% on CursorBench 3.1, costing up to $20 per task
The launch drew criticism for a reversed-axis cost chart
5948.1K4898171.4M
Sentiment
Many users praised Claude Fable 5's record CursorBench performance in Cursor as impressive and enjoyable, while others criticized the evals dashboard for its reversed x-axis, added costs, and billing friction.
Pos
62.2%
Neg
37.8%
101 comments with sentiment.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS226.9KBOOKMARKS81LIKES1.4KRETWEETS27REPLIES43
driss guessous@drisspg
Holy chart crime
Cursor@cursor_ai
Claude Fable 5 is now available in Cursor.
It sets a new state of the art on CursorBench at 72.9%, 8 points above the previous best.
1dViews 226.9KLikes 1.4KBookmarks 81
Theo - t3.gg@theo
@drisspg ?
Top right is good. Bottom left bad. Top left smart but expensive.
This is totally normal?
23hViews 20.3KLikes 373Bookmarks 10
Cursor@cursor_ai
See how Claude Fable 5 compares across every model: http://cursor.com/evals
1dViews 16.6KLikes 78Bookmarks 6
Jacob Beers@JTBeers
@cursor_ai Why is the x axis backwards?
1dViews 4.4KLikes 62
Marco De Donno@stonksandgoats
@drisspg that’s the correct way to show a performance-efficiency curve
1dViews 2.7KLikes 62Bookmarks 1
Sherlock@thesherlocker
@theo @drisspg I think most people assume a cartesian quadrant system.
i.e, both X and Y axes start at 0 at the same point.
23hViews 2.3KLikes 49
kaloszer@kaloszer
@drisspg what's the crime? that price goes the other way? If you read the legend it's all understandable. Maybe the yellow color on white, that's my best guess?
1dViews 4KLikes 49
Every 📧@every
@cursor_ai We've been testing it for a week. Here are the best workflows: https://every.to/vibe-check/anthropic-mythos-our-fable-vibe-check
1dViews 723Likes 9Bookmarks 3
piet@iampiet
@cursor_ai the more things change the more they stay the same
1dViews 2.1KLikes 26
ChickenSamosa@ChickenSamosaa
@drisspg Nah this is the standard for performance-to-cost charts.
Best models are to the top right.
1dViews 1.5KLikes 14Bookmarks 1
Melvin Vivas@DonvitoAI
@cursor_ai $20 per task is crazy
23hViews 1.4KLikes 4Bookmarks 1
Ali Romman@aliromman_
@cursor_ai The benchmarks are insane.
1dViews 2.2KLikes 6Bookmarks 1
Sami Hindi@DevBySami
@cursor_ai Use 5.5 high. Not fair to compare high to medium
1dViews 3.1KLikes 12
mark@r_marked
@drisspg Performance is obviously expected to go up as cost increases. Nobody is interested in the performance as cost approaches infinity, they are interested in how long performance holds up as cost approaches 0. Cost cannot go below 0
22hViews 1.2KLikes 12
Manish Baghel@ManishBaghelz
@drisspg I guess they thought making Composer appear on the rightmost side will be perceived as good.
Instead everyone thinks that thia chart is a crime.
1dViews 3.4KLikes 6
lost in latency@lostinlatencyX
@cursor_ai composer 2.5 is close to fable 5's lowest is what impresses me 😅
1dViews 1.8KLikes 4Bookmarks 1
Leon Lin@LexnLin
@ericzakariasson is low effort good enough? 😭
1dViews 485Likes 5
Nischal@NischalSkanda
@cursor_ai This tells me everything that I needed to know! you guys are the best.
1dViews 504Likes 4Bookmarks 1
Sundaram Kumar Jha@skjtwts
@cursor_ai we need composer 3, Fable 5 cost api cost is too high
1dViews 1KLikes 7
echo.hive@hive_echo
@ericzakariasson I did and it is very impressive!
22hViews 93Likes 5Bookmarks 1
Load more posts