Multi-model setup test: Opus 5.5 + Sol 6.1 wins on cost, DeepSeek V4.1 Flash too weak and needs more back-and-forth
The author compares Opus 5.5 alone and paired with Sonnet 5.5, DeepSeek V4.1 Flash or Sol 6.1, finding Sonnet 5.5 sub-agents a waste of money and Opus+Sol the best cost-quality combo, while DeepSeek V4.1 Flash is too weak and needs more back-and-forth.
As part of working on many projects right now, i've been experimenting with many setups: opus 5.5 alone, opus 5.5+sonnet 5.5, opus 5.5 + deepseek v4.1 and opus 5.5 + sol 6.1 Turns out using sonnet 5.5 sub-agents is giant waste of money! I already felt this but opus 5.5+sol 6.1 are pretty good combo on both $ and quality of output, I think it's cause base models differ quite a lot so they fill in each other's weaknesses. Similiar case for deepseek v4.1 flash but its too weak so more back and forth are necessary, but this model does feel more and more open-ai taste over last few months.. xD