Which models for which mode?

[rangerrick337]

I'm just wondering which models people are using with which mode.
Which models are you using and in what mode?

For example: Boomerang = Gemini 2.5 and Default = Gemini Flash?

Or should the Default also get a smarter model?

[mapringg]

I currently use Gemini 2.5 for all the modes. I did try using Claude for the code mode, but I don't feel like it's worth the extra cost. Personally, I am looking forward to Flash 2.5.

[GreatScottyMac]

I used Gemini 2.5 pro for everything while it was free and it was nearly flawless. Now I'm back to using claude 3.5 sonnet via copilot and VS Code LLM API and for me it's fine for the $10/mo cost.

[BradKML]

What about Qwen/DeepSeek/GLM/Kimi? Any good recommendations?

[wokkieman]

Do you use 3.5 for all modes? That surprised me :)

…

On Fri, Apr 11, 2025, 19:17 GreatScottyMac ***@***.***> wrote: I used Gemini 2.5 pro for everything while it was free and it was nearly flawless. Now I'm back to using claude 3.5 sonnet via copilot and VS Code LLM API and for me it's fine for the $10/mo cost. — Reply to this email directly, view it on GitHub <#47 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA7UH3NE5NWQQMCFE2432TT2Y72JPAVCNFSM6AAAAAB2ZQQSWOVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTEOBQGY4DKMQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.*** com>

[FrancisVarga]

Claude 3-7 but why is it a surprise sofar its the best. just asking tho.

[wokkieman]

Cause each model has their own strength. I would have expected architect to benefit from different strenght then coder

[FrancisVarga]

for now sonnet 3-7 can do both as good and fast any other suggestions?

[SurgeonY]

Using Sonnet 4 for orchestrator and architect and GPT-4.1 for coder and the rest. Going to try Gemini 2.5 pro in place of sonnet.