Cheap and efficient. That's the combo. DeepSeek V4.1 flash delivers real reasoning at a fraction of frontier costs.

Better than GLM5.3 Flash?

Same, with opencode

Same here, changed from minimax m3 and it was a miracle.

Cheers for the tip. I've been thinking about this

What does your stack look like?

Same, Hermes was unexpectedly solid so I've stuck to it and haven't even explored others beside kai9000 on my phone which I also use with deepseek.

hardware?