#Feedback SWE-1.5 + Sonnet 4.5

1 messages · Page 1 of 1 (latest)

lofty hull
fast isle
#

im confused what this is for? is the intent that htis is a more cost effecitve way to plan and execute?

#

i've been using sonnet 4.5 exclusively , since it fails the least and understand context and performs the full execution without much prompt. swe generally needs3x more encouragement, and reprompting to work, wasting more than 3x the credits

lofty hull
#

It’s the new SWE-1.5 model that the team introduced a week ago, working together with Sonnet 4.5 so that both models help each other generate the best possible response for the user.

SWE-1.5 is our latest frontier model, delivering near-SOTA coding performance at unprecedented speed.

slender pumice
#

I cannot see this model, even after updating and restarting Windsurf. I'm on the pro plan. Should I see this / where can I enable it?

lofty hull
slender pumice
#

No, I'm not using Next. Note for the next person: This will only show in Next

harsh hare
#

I've had some good luck using it. There are times where I've seen Sonnet pull SWE out of a loop. I've been using it since it was introduced and the code output has been good.

lean thicket
#

The looping will be fixed in an update coming soon!

thorny path
#

This model seems notorious with looping issues
request id: 9cc2ca89-f18c-4c44-9a83-061c47211361

Also ngl kinda sad that error/failed requests still costs credit

thorny path
#

Can we have Fast Context for SWE-1.5 + Sonnet 4.5?

I've been using Fast Context with Grok Code Fast 1 and it improves overall accuracy by so much.