Whats your workflow, which models?

Discussions
1

It’s hard to keep up with all those new models all the time.

Some weeks ago, I preferred to:

  • plan with Gemini 2.5 pro thinking
  • implement w agent using Sonnet 3.5

Now we have o3, Sonnet 4 (and Gemini) for thinking, and Claude 4 got much better.

I often go directly to agent w Sonnet 4 thinking.

For more complicated stuff, I tried o3 for planning, and Sonnet 4 to implement. o3 feels slow, cause it asks a lot of questions. But that may be an advantage, even. Or Sonnet 4 thinking is good enough?

Interested in your experience!

1 Like
2

Until two weeks ago:

  • 100% Claude 3.5 Sonnet with very specific prompts, rules, .. for both planning and implementation.

Now its Sonnet 4 Thinking Max only, it seems that it just clicked how to write requests that are followed.

1 Like
3

Didn’t mention Opus 4 above, was MAX only. Maybe it’s in the mix with the new PRO unlimited option, not sure.

4

GPT-4.1 anybody, for speed?

5

Saw several threads in forum about reliability issues with GPT-4.1 (also doesnt seem to be that great model from my experience)

1 Like
6

Is Sonnet 4 Thinking (MAX) included with the Pro plan now? Or are you paying the extra fees for MAX?

7

I will have to check that :slight_smile: good point.

Haven’t yet switched to development today.

1 Like
8

Claude 4 without extended thinking is very fast, while being as fast as gpt-4.1

1 Like
9

Max is included in the new plans but note that with any resources used the more you use it the faster you get to the rate limit. Naturally the Ultra plan has higher rate limits.

Therefore less consumption will be with regular models and most/faster consumption will be in Max mode. For most users non-Max will be great usage.

2 Likes