I’ve been wondering here if there is any ETA or roadmap to release new version of the LLM that v0 use. Why this question? Pretty simple… I’ve (I mean, any one of us that has been using v0) been struggling with codebase destruction, random updates / deletion, hallucinations and stuff and this is quite annoying (to not say other thing). And the amount of tokens we waste just clicking in the “Fix with v0” button makes even harder to keep going. One of your competitors (Tempo Labs IA) has an AMAZING feature, they do not charge when their IA mess up.
And I was wondering if this will be here too because this is really amazing!
This question is not to put a target or something, it is just to know what to do next before trying to add more features and be much more annoyed than I’m already am.
Don’t get me wrong, v0 saved me thousand and thousand of hours to get the job done, but I’ve been rewriting this whole codebase aside just for the sake of my peace of mind. The amount of codebase deletion is quite unbearable after the project get really big or have several features.
That’s a seriously great question and critical enhancement. I’ve encountered a few of those build errors, and you almost think you’re stuck in a loop, so it’s only fair for AI errors to offer free fixes.
Releasing the model selector so we can switch between small and medium models with different pricing was the first big step, and that will unlock the power to use lower pricing for error fixes. We have no ETA on this but there is work underway in this direction to reduce the cost of generations whenever possible.
This isn’t quite as easy of a feature as it sounds, and we actually had free error fixes at one point before but some folks found out they could intentionally break the code to get free generations so we had to shut it off
Are you saying that work for cheaper Fix Code generations is coming (e.g: later this year) and Fix Code, will at least automatically use the cheapest model to work? Also, perhaps the first Fix Error is paid, but subsequent should not be. Ultimately this applies only for code which has syntax errors which would not compile correctly. Hopefully an effective tradeoff can be found.
I’m not sure what conclusion the team has exactly settled on here, but at the very least if you are building a project with v0 Large and click Fix in v0, it will definitely not be using the large model for the bug fix. I’m not sure if they’ve shipped that yet but there is active work in this direction
So… hmmm… the vercel’s solution was to penalize every single user because of a few?
But anyway, v0 is still losing in this AI game since their competitors (lovable, bolt, tempo and so on) are way ahead of v0 in the terms of anything. I purchased v0 subscription just because I really though it would be the really best since NextJS is owned by Vercel, so I really thought the LLM would be highly trained with the latest documentation and stuff… but boy oh boy… I was wrong =(
And yeah, I don’t doubt anytime the dev team is working relentlessly to release their newest model version. I just wanted to have an estimation or something to know what to do from now on, since I haven’t touched my project here in the last 2 weeks and I have no idea also when I will be doing something here, but since I already spent some bucks here, I need to know at least when I will be able to spend them more… hhhhmmmm … wisely… You know?
That’s the main reason behind the question. To know when I will be able to keep doing my things here before spending more and more money endlessly…
Basically this was the reason…
And yeah, I know it isn’t easy to release something new (I’m a software engineer who doesn’t know NextJS and frontend, but now, had to learn to improve the code and security), I’m quite aware how hard is and I’m not questioning this anyway.
