Post

Dennis Martinez

I began trying out Claude 3.7 Sonnet in the past two days since I've heard good things since this model came out. It's blown my mind with how much better it is with coding tasks compared to ChatGPT's existing models. Looks like I'm going to switch over my paid subscription soon.

28/3/25, 8:57 am0 reposts1 like

@manmal.bsky.social

IMO the Aider leaderboard is the best indicator for coding performance: aider.chat/docs/leaderb... Claude is #2 (Gemini at #1 is mind blowing btw)

aider.chatAider LLM LeaderboardsQuantitative benchmarks of LLM code editing skill.

Dennis Martinez

Thanks for sharing this! It's pretty impressive how Gemini has seemingly been getting much better lately. I've completely slept on it.

@manmal.bsky.social

It's been not very good up until 3 days or so, you didn't miss much :)