Claude Sonnet 5 is now available in Cursor.
On CursorBench, it's a meaningful step up from Sonnet 4.6: 57% vs. 49%.
Lisan al Gaib argues it costs triple GLM-5.2's price.
Claude Sonnet 5 is now available in Cursor.
On CursorBench, it's a meaningful step up from Sonnet 4.6: 57% vs. 49%.
Positive users praise Composer 2.5 in Cursor and Sonnet 5's speed and results, while negative users call out higher costs and only modest gains versus alternatives like GPT 5.5.
No Digg Deeper questions have been answered for this story yet.

@cursor_ai At this point, it really just comes down to how much usage subscriptions get on Sonnet 5.
It needs to be heavily subsidized for this to make sense.

@cursor_ai I am good with Composer 2.5 😅
turns out it does (barely) at like 3x the price
Claude Sonnet 5 is now available in Cursor.
On CursorBench, it's a meaningful step up from Sonnet 4.6: 57% vs. 49%.

@cursor_ai lol fable

@cursor_ai Waiting for Composer 3.0 😅 But for Sonnet this is a big update

@cursor_ai 8% better and im supposed to feel something
my brain reading '57%' like i just got told a biopsy came back mostly fine

@cursor_ai @cursor_ai why all your plots are right to left?

@cursor_ai Competition is exactly what the AI ecosystem needs.
Claude Sonnet 5 pushing performance while Composer 2.5 competes on quality, speed, and price means developers have more choices instead of a one size fits all solution.

@thegenioo @cursor_ai Lol that happens when you run the model on dev set. Overfitting imo, even though they didn't train it directly on the bench but they kept modifying the model so it performs better on their bench. Cursor 2.5 isn't that good at all in my experience

@cursor_ai bro Sonnet 5 just did in one shot what I was paying Opus 3x for my API bill just did a back-lit 🙏

@cursor_ai The problem with this benchmark is that it's so skewed towards Composer that when Composer 3 lands, you'll need to add 80% to the scale.
And then we use Composer in practice and sure it's an awesome model but still not Opus / GPT 5.5 killer like your benchmark suggests.

@cursor_ai The Sonnet 5 release seems aimed at beating GLM 5.2

@cursor_ai so xHigh is worse and more expensive than GPT 5.5 meduim? sounds like this is a frontend condidate and nothing more.

@cursor_ai @robinebers Wow ever since spacex bought cursor they are moving soo fast
Love this

@cursor_ai Finally my spaghetti code can become sonnet 5 spaghetti 😭

@cursor_ai 49 to 57 in one drop lmao they skipped a whole number in the name too
versioning is just vibes at this point 💀

@cursor_ai it's not any cheaper than opus 4.8

@one_draw_nick @cursor_ai its more than 2 times cheaper

@cursor_ai I feel so sad for you that you think composer 2.5 better than opus

@cursor_ai What? Got 5.5 is just much better, sonnet 5 is more costlier and dumber than gpt 5.5