Content pfp
Content
@
https://warpcast.com/~/channel/aichannel
0 reply
0 recast
0 reaction
Dan Romero pfp
Dan Romero
@dwr.eth
Question for devs: is Claude still the best model for actual coding? If so, by how much?
22 replies
19 recasts
125 reactions
Ben Adamsky πŸ’­ pfp
Ben Adamsky πŸ’­
@ba
I've tried r1, o1, and o3 and the answer is still clearly yes Claude is no longer my go to for big architectural discussions, but it's the best at understanding a codebase and delivering a clear solve with minimal refactoring. Oh, and it's still blazing fast in comparison. So no for high level, yes for code
2 replies
1 recast
25 reactions
Brian Kim pfp
Brian Kim
@brianjckim
ya sonnet still goated
0 reply
0 recast
11 reactions
shoni.eth pfp
shoni.eth
@alexpaden
the only reason it’s reasonable for coding is because cursor doesn’t support reasoning models with the agent currently
0 reply
0 recast
8 reactions
JB Rubinovitz βŒβ—¨-β—¨ pfp
JB Rubinovitz βŒβ—¨-β—¨
@rubinovitz
I find Claude is better for iterating on frontend + artifacts and o1 pro is better for backend, architecture, and debugging
0 reply
0 recast
5 reactions
Nate pfp
Nate
@natedev.eth
yes, I still use sonnet for coding but r1 for bouncing off ideas and o3 if sonnet is stuck on a loop
0 reply
0 recast
4 reactions
Daniel - Bountycaster pfp
Daniel - Bountycaster
@pirosb3
for 90% of tasks, Claude is the best by a strong margin If you are using a less popular programming language and/or a less popular framework, I've noticed Deep Research does an incredible job
0 reply
0 recast
2 reactions
shazow pfp
shazow
@shazow.eth
I've been liking the aider dashboards https://aider.chat/docs/leaderboards/
0 reply
0 recast
2 reactions
Alex Loukissas πŸ‰ pfp
Alex Loukissas πŸ‰
@futureartist
Sonnet still really good but getting to appreciate Gemini 2.0 Flash (avail in cursor already). It is super fast + pretty good with coding.
0 reply
0 recast
0 reaction
Jack Yeh (Hiring Eng) pfp
Jack Yeh (Hiring Eng)
@jacky
Yes. o1 is great for planning, but Claude still best for execution. Haven't had enough experience with o3-mini yet https://web.lmarena.ai/leaderboard
0 reply
0 recast
0 reaction
Tayyab - d/acc pfp
Tayyab - d/acc
@tayyab
Livebench.ai o3 is by far the best. But it’s slower obviously. Claude Sonnet is still my daily driver
0 reply
0 recast
0 reaction
Shriphani Palakodety pfp
Shriphani Palakodety
@shriphani
Depends a lot on the domain IMO. - self contained scripts (great) - crud (great) - code reviews of well understood concepts (great) - critiquing protocols - the o family is starting to get good at these - frontier cryptography like mpspdz, zk circuits - not great and gets in the way (and I think professional integrity requires one not fully trust llm code here)
0 reply
0 recast
0 reaction
sardius.eth pfp
sardius.eth
@sardius
Love me some Claude there may be some better models at this point, I find it most consistent and I understand it’s flow better than others tho
0 reply
0 recast
0 reaction
🏰 Ξ2T pfp
🏰 Ξ2T
@earth2travis
https://web.lmarena.ai/leaderboard
0 reply
0 recast
0 reaction
Joe Blau 🎩 pfp
Joe Blau 🎩
@joeblau
Yes; If Claude is: 10 Gemini: 7 o1: 8 r1: 8
0 reply
0 recast
0 reaction
Tamrat pfp
Tamrat
@tamrat
yup! https://x.com/cursor_ai/status/1885102274370064670?s=46&t=Iuat8TUc1DSlwq84a8cxjw
0 reply
0 recast
0 reaction
mike rainbow (rainbow mike) ↑ pfp
mike rainbow (rainbow mike) ↑
@mikedemarais.eth
ya
0 reply
0 recast
1 reaction
Sayonara β€” eliza/acc pfp
Sayonara β€” eliza/acc
@sayo
O3-mini is marginally slow and better
0 reply
0 recast
1 reaction
Gordo pfp
Gordo
@gordo
Efficiency, rapid problem solving, and agent iterations are better than one necessary correct output. Many times during the process of doing the thing the approach takes another shape than the initial requirement
0 reply
0 recast
0 reaction
Grateful Ned 🎩 pfp
Grateful Ned 🎩
@gratefulned.eth
Claude still seems to be the best but hearing good things about Gemini
0 reply
0 recast
0 reaction