I've been using Claude Sonnet 3.7 as my primary model and Haiku 3.5 for faster tasks in Claude Code, and it seems to be working really well. I'm curious, what models are you all currently using as your main and fast models?
6 Answers
I’ve always preferred Sonnet over Opus. Sonnet 4 has really become my go-to!
Right now, I'm using Claude Sonnet 4 and Haiku 3.5. They suit my needs pretty well!
I’ve switched to Claude 4.0 in ClaudeCode. It integrates Sonnet and Haiku seamlessly for whatever I throw at it. I tried Opus for fun, but honestly, I haven’t found a necessity for it. It handles more complex tasks, but I think my tasks are just too simple right now!
It's funny how people were bashing Sonnet 3.7 and swearing by Sonnet 3.5, but now a lot of us have adjusted. I think each version brings some improvements that might need us to tweak our prompts a bit. Personally, I'm mostly sticking with Sonnet 3.7 too! Also, Haiku works great for quick tasks or command calls.
Exactly, if we had to constantly relearn to play an instrument, no one would pick it up!
Wait, you can actually use Haiku 3.5 in Claude Code? That's news to me!
Yeah, you can set it up with a main model and a fast model. The default fast model is Haiku 3.5, but availability depends on your provider—I use AWS Bedrock.
Opus 4 with Gemini 2.5 pro is where I'm at right now!
Really interesting choice! What does your process look like with that setup?

Imagine if hammers had the same level of updates as LLMs!