See Anthropic’s announcement here.
Super excited to see how performance of Replit Agent & Replit Assistant improves on the new 3.7 model!
2 Likes
great news! when do you think we can start playing around with it within replit?
1 Like
Eh, more excited about Grok 3. It’s crushing it right now. I’ve abandoned the use of all other AI.
I’ve been using 3.7 in Cursor (on my replit) and it’s been able to fix and make changes that Replit with sonnet 3.5 couldn’t. The thinking model looks very good (if a bit verbose and slower)
Sonnet 3.7 is looking better than Grok in the stats, plus Claude have the best and most reliable model - now they’ve added thinking models to the mix they’re going to be practically very hard to beat.
4 Likes
Just got access to Claude Code today and I’m super impressed. $1.37 to do a total overhaul of my web app’s UX, fixing multiple bugs that I’ve already spent at least $5 trying to fix with Replit on 3.5 Sonnet. Really hoping they bring Claude 3.7 Sonnet with the thinking model into Replit soon!
Claude 3.7 sonnet is active for Assistant now and v2 Agent uses it extensively.
just activate “Explorer” on your account by clicking your profile picture in the top right → Click “Account” → Scroll to “Roles” → Toggle on “Explorer”
2 Likes
Oh hell yeah! Did not expect such a quick turnaround, that’s great news! Thanks @kody-replit! Gonna go start try it out right now.
1 Like
Hey what happened to 3.7? It disappeared.
Thank you. It was on 3.7 before - never even seen the settings page before.
Got it fixed.
Yup we tested defaulting to it, got negative feedback so defaulting to 3.5 until we get 3.7 more reliable on assistant. Agent v2 was a total rewrite to use 3.7 more effectively, assistant still had a lot of the 3.5 specific infra around it so the flip to 3.7 was a downgrade.
I find it most effective with 3.7 plus this custom prompt:
1 Like
Just implemented this. Excited to try it out. Thanks Kody!
1 Like