A couple of months ago, I did some extensive work with Pinegrow on a few static Tailwind sites and used Mr Pinecone with Openrouter in auto mode. Looking at the Openrouter logs, Quick was using Gemini Flash and Smart was using Gemini Pro. I tended to use Smart for initial project briefs and more creative work, then rely on Quick for more targeted edits. Smart was a lot more expensive in tokens and also went rogue at times, making unrequested changes.
Since I did that work, Gemini Flash has progressed to version 3.8 while Pro has remained stuck on 3.1. I plan to start on some new projects and wondered if Smart is now defaulting to 3.8 Flow and any recommendations?
I have a Gemini Pro subscription, so I just asked there and it suggested using Gemini Flash 3.8 for both Quick and Smart, setting them manually.
Thanks, I guess I’ll give it another try with the new projects and see how it goes, though it would be good to hear some feedback from the devs on their current experience and recommendations.
Curiously, I received a newsletter from Openrouter earlier saying that requests are load balanced on price by default, but that doesn’t necessarily reflect ability. Your local model solution might be the right way forward for all of this.
I have never tried Deepseek, but from what I hear, it is super economical to run, so if it works that is definitely worth trying. It would almost be good if Pinegrow could show which model is in active use when auto is selected for Openrouter. If it is jumping around between models they all have their pros and cons in theory, so that might cause inconsistency in the design.
I have been stung financially by Claude as well previously with other tasks outside Pinegrow, so I am always hesitant to trust it.
Just now, I opened Pinegrow 10 for the first time and figured I would take a look at the AI models available through Openrouter, but all the listed options are curiously outdated, as shown in the attachment. If any of the Pinegrow developers see this it would be helpful to have more current choices.
The most popular model right now on the Openrouter site is called Space Bunny Alpha, which is literally free. Next comes DeepSeek: DeepSeek V4.1 Flash.
EDIT: I have added flash 3.8 through the Openrouter additional models field, so that shows now, but it would still be good to have some feedback from the devs.
EDIT 2: After further testing, Smart is still defaulting to Gemini Pro 3.1, which seems to be nowhere near as good as flash 3.8 and costs about 10x more to run. For the time being I will keep it fixed on flash 3.8 through Openrouter.
Late to that party I know.
