[▲ Vercel Community](/) · [Categories](/categories) · [Latest](/latest) · [Top](/top) · [Live](/live) [v0](/c/v0/59) # Agent mode feedback thread 2385 views · 232 likes · 80 posts Jacob Paris (@jacobparis) · 2025-08-11 · ♥ 6 v0 has just launched Agent mode which adds - task list creation - queued messages - looking up docs for libraries you mention - automatic model selection for tasks and subtasks https://v0.app/docs/agentic-features The automatic model selection means v0 can use a different model for reasoning than it uses for planning and again a different model for actually generating the code. For example it might pick GPT 5 for planning or design tasks and then switch to v0-1.5-md to reliably output the code For this reason the model selector and the GPT-5 switch have been removed Watch it in action: https://community.vercel.com/t/-/18060/1 I'm making this thread to consolidate early feedback on the feature so the v0 team can iterate on it quickly Please be constructive in the comments - if you are hitting bugs please provide a chat ID and supporting evidence so we can try to reproduce it - if agent mode as a workflow doesn't feel suitable for you, please explain more about your use-cases and reasoning so it can be taken into account for future planning iterations Syazwan (@mohamadpaharudin-477) · 2025-08-11 · ♥ 17 Hi Vercel Support, I’ve noticed that my V0 workspace has been migrated from `v0.dev` to `v0.app` and now defaults to the new *Agentic* mode. While I understand the intent to streamline the experience, this change has removed the ability to manually choose the model (e.g., GPT-4o, GPT-5o) for my prompts. For my workflow, manual model control is critical. I often need to: * Compare outputs from different models for the same prompt. * Ensure deterministic behavior in production prototypes. * Optimize for cost or speed depending on the task. With Agentic mode always enabled, I have less control over output style and model capabilities, which impacts my ability to deliver precise results for clients and maintain consistency across revisions. **Request:** Please provide an option to disable Agentic mode in my account or for specific projects, restoring the previous manual model selection capability. If a global toggle isn’t possible, even a hidden or developer setting would be extremely helpful. I appreciate the work going into improving V0 and I understand the vision for Agentic workflows, but for advanced users like myself, the loss of this control is a significant regression. Looking forward to your guidance on whether this is possible or if a workaround exists. Thank you, Syazwan Skudy (@skudy) · 2025-08-11 · ♥ 11 I’ve tried multiple times to use GPT-5, but the Agent mode doesn’t seem to be utilizing GPT-5 appropriately. It feels like they’re testing on paying users, which is very uncomfortable. I can tolerate frequent UI changes, but since this is a development tool, changes in the development workflow are extremely confusing. Mtspoelstra (@mtspoelstra) · 2025-08-11 · ♥ 9 I would also appreciate a way to disable agent mode and use GPT-5 again please. Richard 2105 (@richard-2105) · 2025-08-11 · ♥ 3 The agent seems to take a lot more liberties with my prompts as I try to iterate on a design. So it seems like it doesn’t do anything that I’m actually asking. You can take a look at my project for troubleshooting. If possible, can you revert this agent for iteration? I do like what it’s doing. It does have a good design taste. Kaintetro 7346 (@kaintetro-7346) · 2025-08-11 · ♥ 9 Hey guys, I really like v0 and I don’t want to switch tools, but the agent mode is really taking away my control over the budget and the projects. I’d like to have the option to switch it off or exit it. reoring (@reoring) · 2025-08-11 · ♥ 6 Hi Vercel team, Thanks for shipping fast—Agent mode is an interesting direction. However, after the migration to v0.app, I’m seeing a significant drop in output quality for my workflow: the agent’s results are noticeably weaker in **visual design quality** and **content depth**, and it takes too many liberties with my prompts. This makes iterative design harder and less predictable. For my day-to-day, I need tighter control and reproducibility. **Requests** * **Allow disabling Agent mode** (workspace / project / thread level). * **Restore manual model selection** so I can optimize for cost/speed and compare outputs across models. Agentic workflows can be valuable, but removing the non-agent path is a regression for advanced users who need precision and consistency. I’m happy to provide example chats or project IDs to help you reproduce the issues. Thank you. lukek (@lukek) · 2025-08-12 · ♥ 1 I haven’t tried it out, but self-picking seems better because I’d definitely much rather have consistently successfully executed prompts at the expense of a few more tokens, but unsuccessful prompts chew tokens anyway so it’s sounds like a no brainer. I’ll try it out later today 🔥 Itmanner (@itmanner) · 2025-08-12 · ♥ 6 You are so stupid. Did our money come from the wind? Why do you use our cash to test for you? Niaz (@niazique-4279) · 2025-08-12 · ♥ 4 Guys, the agent isn't working well. I feel like I'm using GPT-3.5 because it makes silly mistakes on very simple requests. I hope you’ll switch back to manually selecting the LLM ASAP. selena (@selena-starbridge) · 2025-08-12 · ♥ 7 To the V0/Vercel team, I hope this finds you well. Before the migration, I could choose which model to run, such as GPT-4o or GPT-5o. This allowed me to compare outputs for the same prompt, ensure deterministic behavior for prototypes, and make tradeoffs between cost and speed. With Agent Mode always enabled, I no longer have that control. The system decides for me, which introduces variability in style, fidelity, and capabilities from run to run. That unpredictability slows iteration, makes regression testing more difficult, and increases the risk of introducing inconsistencies into a live design. Since the change, I have also seen a drop in quality for my design workflows. Visual outputs feel less polished, content depth is reduced, and the agent often diverges from the specific instructions in my prompts. This forces me to spend time reworking outputs rather than refining them, which breaks the momentum that is critical in iterative design. My use case requires **stability** and **control**. I need to lock a specific model for the duration of a project, maintain a consistent output style across revisions, and ensure the tool does not alter core behaviors without my direction. If Agent Mode could be made optional, even at the project level, it would allow me to continue working in V0 with the confidence that my design process and delivery quality will not be compromised. BestCodes (@bestcodes) · 2025-08-12 · ♥ 3 My feedback: I love it! It feels way more powerful and like it integrates with things so much better. Keep up the good work, Vercel :heart: LeoGPrime (@leogamiprime-1924) · 2025-08-12 I’ve personally had pretty good luck with agents across multiple projects. It doesn’t feel worse, it still feels fine doing incremental work. You can still confine behavior you don’t want it to do via prompt. I feel like it has just cut down on errors a bit by reviewing it’s own work (Although it’s not close to perfect yet) Just giving my 2 cents. However, I did notice once agents switched, projects that worked just fine completely broke. In my case, it was just a tiny change in the middleware.ts script located inside /lib/supabase. Just throwing this in here as a bonus, although it might not be appropriate since it’s not related to agents, but dragging a .rar file doesn’t properly port a project over properly anymore. (For some weeks now) It loses some directory/file structure integrity. However, oddly enough, if you re-zip it as a .zip instead of a .rar, it preserves the directory integrity and properly ports over projects. Amsend (@armsend) · 2025-08-12 · ♥ 4 Greetings, Shipping this huge change to paying users and taking the money that we spent and charging us for the prompts which i’ve noticed cost more now was a very unwise decision. The Agent is making alot of mistakes and so far i’ve brunt through the gifted credits in less than 2 hours and now it will do the same to the money i paid. add an option for us to revert back and disable agent mode. thank you. Adallbertong 6490 (@adallbertong-6490) · 2025-08-12 · ♥ 5 Before the update, everything I asked for, it was able to do. Now, with the update, I’m not getting the same results. Besides that, it keeps repeating the process over and over, taking a long time to deliver the result — something that used to be quick. A simple change in my code took more than 10 minutes, when before it took only 2 to 3 minutes, not to mention how many times it repeated the process, just wasting my credits. On the very first run, it went into a loop on a simple task that I used to complete without any trouble, and it almost maxed out my tokens — this is unacceptable. Also, they are forcing the use of agents, and I want to be able to disable that. I’m hating this update, and I think I want my money back. Arian 2020 (@arian-2020) · 2025-08-12 · ♥ 4 Hi, I mean this is okay but having the option to choose between models its better like leave us some `Control...` I dont like the agent mode… choosing between models was better and even more helpful, models were able to generate over 3k code in only 1 file, now this one cant even write 1.5k code, and it laggs a lot. Panostampa (@panostampa) · 2025-08-12 · ♥ 6 you cant choose the level of ai no more , you are stuck with agent mode which eats up your credits and does nothing at all , horrible update sirube (@sirubedon) · 2025-08-12 · ♥ 2 Agent Mode is very convenient. It’s convenient, but I want to be able to choose the model it runs on. I can only use md all the time. I’d like to be able to use higher-quality models like lg or GPT-5. That’s because md might destroy what I’ve created or misunderstand my intent. I’m willing to accept some credit consumption, so I’d like it reverted. Daniel (@analisedadosforever) · 2025-08-12 Essa nova atualização pra agent é horrível, você pede uma coisa simples e ela faz outras não pediu, estava Prestes a acabar meu projeto pra rodar e agora vou perder dias tentando, tem que ter a opção de desativar. Ikkelucky (@ikkelucky) · 2025-08-12 · ♥ 2  So I know the screenshot is kind of not showing much but its actually showing a lot. I got this response like at least 4 times, consuming at least 5usd worth of tokens (1.5 more less every time) and in the end it got stuck every single time. It does not give you a button to carry on, it does not give you back the amount spent, it does not even leave the work done so far. With previous models in which I could choose from, working with tasks that I used to work with will get challenging to say the least. I am thankful though since I got some credits to try it, so its actually not that much of a complaint but i would like to keep working with v0 and Vercel in the way it used to be before, or at least to remodel the agent in a way that it actually adds the code it was working on! To clarify, i asked to create 10 replicas of a form and success page. 20 components might seem excessive, but they are all replicas from each other and it was something that the previous models could handle easily arriving at about 10-15 components created at the same time, without spending as much tokens, credits and without making mistakes (or not even delivering the code created so far!) If there is any other way to create code with this model in a way that it wont stop working and implement it? I liked some of the changes, but it could be problematic on a potential bigger scale (in my case, that i handle around 10 corporate apps/accounts). This might drive me (and possible much many others) to pass on some of these new features if there is no possibility to effectively consume the product that self deletes or stops the code that was requested (and paid for) Alexundo (@alexundo) · 2025-08-12 · ♥ 2 So, big fan of v0, I have been using it extensively since last year. It is great, especially for someone that comes from design. It always seemed to me that v0 was the superior “designer” out there. Now adding the Agent mode is a welcomed addition, it has great potential and clearly this is the way. But. As it stands something is terribly wrong, either agents configuration, model, whatever. The main issue is that the agent mode is much more costly, which would be fine if the output was high quality. But it’s not, I have completely burned through the gifted credits yesterday and today. Today especially was the worst experience, burned the $20 and barley got what I asked for, in fact is about 50% of what I asked extensively even providing documentation and explicit requests. Design is the worst it has ever been, but also it lacks any kind of project awareness, always defaulting to basic shadcn components instead of using the custom global components already present in the project. I was used to stop explaining basic UI or Development good practices as it would at least cover that, but now it just seems like I’m working with the older models, where every steps was a chore but we didn’t have anything better. Bottom line: Agent mode = cost more provides less. Not a great combination. While I believe it is a necessary step in the right direction, it needs a bit more time in the oven. Contato 4630 (@contato-4630) · 2025-08-12 · ♥ 8 This update has negatively impacted a project that was in progress. The agent mode is far inferior to what it was before; now I have to spend $10 in credits to get a result that I would have gotten for $0.10 with the previous version. This update needs to be reverted urgently. Vitalysp1 (@vitalysp1) · 2025-08-12 · ♥ 5 Please add the option to switch between modes. The current Agent mode feels unfinished and inefficient. Even if it were fully functional, I would still want the flexibility to control and use other models. The ability to edit specific elements in real time was an excellent feature and should not have been removed. I think this community enjoys autonomy and this clearly takes it away. I hope this doesnt turn into squarespacish where stay at home moms will be “vibe coding” their next MLM offering Tomas (@tomas) · 2025-08-12 Not sure if this is related, but preview is broken for me (testing a small app in vue.js)  riza (@hasriza) · 2025-08-12 · ♥ 1 **Pros:** * Performs deep code review and finds issues that would otherwise go unnoticed * Less overhead on our part to decide what would be suitable for which task * Ability to reproduce design from another website sounds amazing, haven't tried it though * Good suggestions on how to move forward with the project **Cons:** * Works on changes that break the system (instead it could simply suggest and then proceed if user confirms) * Creates duplicate files that already exist and on top of it, copies the same code... :neutral_face: * Initially creates a separate project referencing the existing project, probably a version difference issue (creates app at root instead of src). * Burns through credit a lot more due to lack of choice in cases where we actually know which model would have been better to restrict output size * Extensive research material in the response, everything in it is important but would be better if it was shortened **Suggestions:** * Respect users' opinion to have a model selector and an agent toggler that would allow user to control the output and whether the process should be totally automated or semi-manual. * Context window seems to be smaller which possible ignores certain files and duplicates them * Improve the agent's code review knowledge by having it regressively check if the file changes and code it has planned align with the user's request while also ensuring that it hasn't already been implemented in some other file with same or different name across the project * Add buttons in the response window that allows reverting file or cancelling changes Mtspoelstra (@mtspoelstra) · 2025-08-12 · ♥ 7 I’ve been using v0 for a while now and have really valued how it makes developing my projects faster, easier, and just generally more productive. When the option to select GPT-5 became available, the jump in capability was remarkable and it completely transformed what I could accomplish in a short time and with minimal effort. My experience with agent mode has been the complete opposite - this update is breaking my workflow and making it impossible to continue my project with v0. It does not produce what I ask for, even with explicit, clear instructions. It frequently claims to have implemented changes it hasn’t, makes significantly more errors, ignores best practices that GPT-5 followed without much guidance, takes far longer to produce results, delivers much poorer overall quality in the UI and in the code itself, and costs more in both time and money. At this point, v0 is simply not producing the quality or accuracy I was getting with GPT-5, or even earlier versions, and I cannot move forward with my work under these conditions. It is wasting my time and is putting me at risk of missing deadlines. I request an immediate option to disable agent mode and select GPT-5 directly again. If this is not resolved, I am a day or two away from having to migrate to another platform and request a refund. Georgeaffield (@georgeaffield) · 2025-08-12 · ♥ 6 Guys, PLEASE REVERT BACK!!! This Agent mode is god awful! Your previous version of being able to select small, medium, large ect was soo good! It literally shaved about 3 weeks of dev time for a recent UI implementation needed on a project. Agent mode on the other hand has been the complete opposite, its like going back to 2022 again where the it takes so much time to prompt and nothing changes. Now I am spending more time trying to get the agent to do things when it would be far easier for me to code it. I am finding that I am having to step in and make far more manual changes myself too. So far, really poor experience, you guys had it nailed Uğur (@ugurkorkmazyurek) · 2025-08-12 · ♥ 7 I thought the problem was with me. The previous version understood everything better and did everything perfectly. But this update can’t do anything right. It’s a complete disaster! Adi Darachi (@adi-darachi) · 2025-08-12 · ♥ 3 tl:dr; Agent mode does not work. after about 40 iterations trying to do simple things I could previously do easily I am forced to cancel the subscription and move to a more stable tool. I would like to contribute and back other comments in this thread. In the last two days. since the release of Agent mode I was not able to make a single prompt to do my I intend it to do. That’s after about 700 prompts to the same project which worked decently and was providing me great value. At this moment, I am attempting to migrate to another tool as it simply does not work. In addtion, it now burns through credits like crazy, In about 1h I was able to spend what usually I spend in 2-3 days. I used to spend about 3-5 dollars a day (5-6h) while now I was able to spend 8$ in less than an hour Unfortunately I just canceled the subscription as I don’t see a way for me to continue developing using the tool. I cant stress it enough, It does not work, at all! Cyril62880 (@cyril62880) · 2025-08-12 · ♥ 2 The idea is good, but it was much better before this update, I have been wasting a huge amount of time on a project for 2 days, the model takes too much initiative on code changes, it makes incomplete code, I have a response that lasted more than 5 minutes and took 2€ from my wallet, for a code reduced by 80% (which does not work). Horrible for professional use, I request a return to v0.dev, where I would be forced to change the AI model. You are killing your own AI and it’s very very sad Ross (@ross-7453) · 2025-08-12 · ♥ 2 I miss GPT-5, it solved bugs that all 3 other llm modes weren’t able to solve for 48 hours. For a brief period, I was able to make real progress. Now with agent mode, it’s back to being dumb and making the same mistakes over and over again. :/ Woker (@cuentas-3945) · 2025-08-13 · ♥ 1 The idea of Agent Mode is great, and I appreciate that you are working on making v0 better and better. However, this update needed more work before being released. First of all, the quality of the responses has suffered a major downgrade. I used to use v0 to develop Django templates (HTML), and it was perfect. I would send a file as a sample and specify that I needed the same thing done on another file but with different data. The response was great, and with a few small iterations, I had the page ready for production. Now, the output is a disaster — it breaks the format, invents new blocks that don’t exist, and fails to follow the design pattern of the sample (both in code and UI design). Also, Agent Mode tries to “review” the work. With a new prompt, it reviews it once and shows me the code, but if I ask for a fix, it reviews the same thing five times and writes the same code five times. This also multiplies the billing by five. Luckily, I’m using gifted credits, but for any prompt that would normally cost 15–20 cents, it now costs 60–80 cents (though this issue doesn’t happen with new chats — new chat charges are correct). Please roll back this change. I can’t use v0 anymore with this new version. Boxler (@boxler) · 2025-08-13 · ♥ 3 the new agent mode is so bad… Sezotove (@sezotove) · 2025-08-13 · ♥ 4 I agree with pretty much everyone here. This needs to be rolled back or give us the option. In 2 days I’ve exhausted my monthly credits with the new “model” making tons of changes I never requested or hinted at. I’ve had to force it to revert meaning it’s costing me more. It’s a relatively poor role out as the intended features like checking it’s own work…clearly not functioning properly. lukek (@lukek) · 2025-08-13 · ♥ 1 When I send a message I end up getting put back to the top of the chat and needing to scroll down to see the latest version. tomaspozo (@tomaspozo) · 2025-08-13 · ♥ 3 I keep getting this error over and over again… It fixes, and 2 prompts later, this pops up… very annoying!  Arian 2020 (@arian-2020) · 2025-08-13 · ♥ 2 what a beautiful agent ? dont u think so ? 4 times in a row same message haha!! not worth it at all !!! Clemensrau1337 (@clemensrau1337) · 2025-08-13 · ♥ 6 This is the worst update ever made. All my projects cannot be continued from now! this is getting worse!!! Agent mode is destroying all projects and dont get any ideas how to fix my problems!!! I already cancelled my subscription Vani (@rahulgautama97-2599) · 2025-08-13 · ♥ 4 I don’t want to sound rude so I’ll say it subtly. Let me change back to gpt-5 instead of Agent mode. Why? - contextual aweness zero - hallusination on every other prompt - expensive on every prompt - silly mistakes making my head hurt trying to fix them manually when gpt-5 was clearly better - multiple refreshes needed before I see the result lukek (@lukek) · 2025-08-12 · ♥ 2 Ah okay. So they’re purely to get some feedback on the new Agentic AI. I happened to not use it too much today but what I noticed while using it is: - it attempts to solve the issue in 1 message rather than me having to reiterate myself. - it may miscode other parts of the project which causes other error messages. - it chews through a lot of tokens which is okay but it has to mean the problem gets fixed without miscoding the rest of the project. - more intelligent and remembers better. I’ll also add that finishing a message with “Remember to keep everything else in the project the same” works a lot of the time in Agentic AI miscoding the other parts of the project. Mtspoelstra (@mtspoelstra) · 2025-08-13 · ♥ 7 Will the option to disable agent mode and choose the model I want, including GPT-5 (as it was before the update), be reinstated? If so, when? I cannot keep sitting in this in-between state, unable to work on my project, waiting for someone from Vercel to respond to all the negative feedback on agent mode, and hoping you will actually listen and give us the option to choose again. I do not want, and cannot wait for, a “better” version of agent mode. I simply want the ability to turn it off and use GPT-5 again - which was genuinely phenomenal. Right now, it seems like you’ve launched a product too early, with too many errors to be functional, and you are forcing paying customers to act as test subjects. If you want testing, make it opt-in - but give us the option to turn it off so we can actually get work done. I do not appreciate being tested on as a paying user with a broken product I cannot rely on for my work. If this continues, I will have to cancel my subscription and move to another platform. Ainesh Ravi (@ainesh-ravi) · 2025-08-13 · ♥ 3 Ditto a lot of the comments here - agent mode is not good. I’ve immediately found myself going directly to GPT5 asking a specific problem and then copy and pasting in my code. Please revert this back or make agent mode much better. It’s almost made vercel not usable. Brianshiitake (@brianshiitake) · 2025-08-14 · ♥ 3 Please, guys, this change so abruptly was simply NOT it. bring back model selection, or at least let me pay more money to get it back??? Tlfla (@tlfla) · 2025-08-14 · ♥ 2 Like others here, I'm finding agent mode challenging to work with on existing projects. **S**pecific issues I've encountered: * File corruption during targeted fixes - Asked to fix a rate limit boolean check in one function, agent deleted the entire 200+ line API route file, leaving just one import statement. Had to manually restore from an earlier version. * Waypoint restore not functioning - When trying to roll back after breaking changes, the waypoint feature didn't actually restore the previous state * Scope creep on changes - Simple requests affect unrelated files (asked for a UI update, it modified middleware error handling) I appreciate that agent mode is trying to be more comprehensive, but for those of us with functioning projects that just need targeted fixes, the old model selection was more reliable. Even just an opt-out toggle while you refine agent mode would help us keep working… Not trying to pile on, v0 has been great for my project overall. Just hoping this feedback helps improve the experience for everyone. lukek (@lukek) · 2025-08-14 I’ve noticed a drop in Agentic AI’s intelligence the past 24 hours or so. When I say intelligence what I mean is how good it is at solving specific problems I ask it to. Clemensrau1337 (@clemensrau1337) · 2025-08-14 I finally cancelled my subscription. I cannot use it as before, so I dont Need it any more. Adeola (@adeleye-adeola) · 2025-08-14 · ♥ 1 PLEASE CAN WE HAVE A WAY TO SWITCH MODELS, THIS AGENTIC MODE IS STRESSING ME OUT, REALLY WELL. The agentic ai is not giving good results, like it fraustrating, please give us chance to choose models, it really delaying me Julien (@juliengobbi-5344) · 2025-08-14 Hey! I’ve been using it for a week and the results were awesome, but since the new version it’s been a mess… I’ve tried 20 prompts in a row and none were satisfying. What happened? Pg 7046 (@pg-7046) · 2025-08-14 · ♥ 1 Worst update ever. Previous models delivered design at the level of seasoned UI professional, current model - at the level of a junior (or even intern) :angry: Mike (@mikefieldcs-1978) · 2025-08-14 · ♥ 1 @juliengobbi-5344 They introduced something called “Agentic Mode,” but it has completely broken the system. I can no longer make any file edits or add new features, it fails to understand any of my commands, cannot read the provided .php file, reads it incompletely, and when delivering, although the code is 2,000 lines, it repeatedly gets stuck around line 600. The v0.app platform is in a very poor state. When I explain my request in detail, it not only produces results that are entirely unrelated to what I asked, but also severely damages my existing PHP code to the point that I could not recover my file. For example, if I request to add a modal—describing precisely how I want it—it ends up breaking the site header, inserting numerous unwanted elements into the center of the site, and making arbitrary, illogical, and highly disruptive revisions to the overall layout without any instruction from me. It even adds an excessive amount of unnecessary elements along the right side of the site and all the way down to the bottom. In short, it completely disrupts the site’s functionality and structure in a way that is far from what I requested. Since no representative from the V0 team has provided any information on why this is happening or when it will be fixed, we are left surprised and waiting—hoping they will eventually explain the situation and inform us of when it will be resolved. Rusty 7438 (@rusty-7438) · 2025-08-15 · ♥ 4 **Move Over, The v0 AI Agent is Taking The Wheel!:** Death of the v0-1.5-sm model, squashing the free tier and the cost of autonomy I want to echo the frustration I've seen regarding the forced migration to the AI Agent mode. I know this may seem like exposition, but I wanted to share my experience in case it helps others, as the situation wasn't clear to me until after the fact. But first, I’ve loved v0 in the early stages. The pricing models, as they evolved, seemed to be tracking well with output produced from the ai copilot experience. Even with the sm,md,lg model breakdown, the pricing seemed to align. I could invoke sm model prompts and tweak designs to a satisfying end result and it felt like I was truly working collaboratively with the AI. I was able to choose the amount of assistance needed, and the cost seemed to be compatible. Jump to the post AI Agent experience and it feels like the dynamic has flipped. In the below example. I asked for the div background color for these badges to be modified. A fairly simple request that in the past would cost \~.05 credits, but for the AI Agent cost .39 credits an increase of 680% and nearly all of that cost (.38 credits) are input credits. Meaning a credit pool that used to allow for extensive collaboration now can disappear in just a few prompts and the $5.00 free credit plan So whats going on?  Gone is the cost effective sm model and the ability to manually select different models (ie. manage your prompt cost). With each prompt the AI agent model can [u]choose[/u] to plan, research, build, and debug. Each of these sub actions, as well as any follow up prompts the Agent invokes on it’s own, draws from your \[input\] credits and is performed unchecked by the user. From an accuracy and efficiency perspective, I’ve encountered more looping conversation, particularly around debugging where the AI ultimately couldn’t fix the problem and ends up invoking extra steps to undo what it had just tried or will undo code changes that I have manually committed (at my cost). I also feel like the collaborative experience has suffered because my focus has shifted to prioritizing bulk step prompts that I’ve tested in other AI services in an attempt to maximize individual prompt payoff and reduce repeat debugging. **TRDL**: In essences my experience with the v0 AI Agent has meant, increased cost, increased unpredictability for both prompt cost and output, loss of control and deceased collaborative intent. All of which has lead to increased frustration in my cost to output experience perception So far I’ve rattled on about my experience, my own subjective observations, so lets take a look from another perspective. Here’s some numbers, this was alarming to realize. I hope that something will change.  - Jonasfh2020 2058 (@jonasfh2020-2058) · 2025-08-15 This this this this this 📢📢📢📢 Gabriel - DF (@xnwaltergabriel-1118) · 2025-08-15 The performance of Agent mode is interesting, but there’s a problem. Personally, I used SM to build the base, which saved quite a lot, and MD for design and corrections — here there’s no way to choose between one or the other. The agent makes fewer mistakes, but the quality has dropped — both in design quality and in resolutions. When it comes to tasks involving mobile design corrections, it can’t manage to keep the same template as before. Although, I do admit that the number of errors has gone down. The cost remains similar to MD, but in some file types where I compared an earlier edit to the same edit again, it actually turned out higher, from $0.08 to $0.14. Contato 4630 (@contato-4630) · 2025-08-15 · ♥ 2 I understand that this Agent feature is a test. But I want you to understand that this is harming many projects. I'm not managing to create a simple landing page with the agent mode. I'm having to change the code manually, and this doesn't make sense. I wanted at least an answer from you if the agent mode will continue forever or if you will revert the situation. I need to act quickly because I have projects with tight deadlines. @jacobparis Brandon Haynes (@brandon-5778) · 2025-08-15 There HAS to be an ability to revert back to the old v0.dev without agent mode - it’s significantly worse now, I can’t choose model, the agent keeping flipping thinking too much and overthinking and making too many small changes, and performance in the UI has gone really sluggish too! PLEASE let us revert back to how it was - this agent sucks deeply. OH - and there are </merged_code> issues, it keeps appending the tag at the bottom of changed files and doesnt remove it, causing project to break. meshia (@kinkofa) · 2025-08-15 The potential here is promising, but in its current form it’s slowing me down. **Main Issues** * Very simple messages are costing $2–$3.35, much higher than what I’m used to. * The agent pauses frequently without explanation. * Says it’s resolved issues when nothing has changed; console errors keep repeating. * I’m spending credits on repeated fix attempts that don’t actually fix the problem. * It doesn’t always follow explicit directions. **Workflow Limitation** I want to be able to edit design and copy directly — without generating a message — but that doesn’t work consistently across the app. Having to run everything through the agent slows iteration and increases costs. **Suggestions** * Lower credit burn for low-complexity actions. * Improve reliability and validation of “fix” messages before marking tasks as complete. * Make direct editing of design and copy possible in all parts of the app. * Add better error handling so the same bug isn’t “fixed” multiple times without actual change. Thank you for the $80 in credits over the past few days. I had stopped prototyping the way I used to after the May/June pricing change. The $80 in extra credits over the past few days has been a huge boost and it’s brought back my energy to build. That said, I’m worried about how quickly credit will burn if agent mode is the only option for these tasks. Would love to hear back from the team on these issues. Vlbxrtttt10 (@vlbxrtttt10) · 2025-08-15 Hahaha, well, it repeats the process 5 or even 8 times for me… what a horrible update the agent has! Tzuyi 1494 (@tzuyiyu-1494) · 2025-08-16 We need other models. PLEASE CAN WE HAVE A WAY TO SWITCH MODELS, THIS AGENT MODE IS NOT READY AT ALL!! lukek (@lukek) · 2025-08-16 In my personal experience I haven’t been able to solve just 1 problem over the last few days and right now I’m just hoping more updates come to fix whatever the weakness in the Agentic AI is. The devs can revert back to the old ways, but I think there just needs to be simple my better at solving our problems and if it does that consistently nobody will care if it chews a couple more tokens here and there. If I keep sending prompts to try solve 1 problem and Agentic AI can’t solve it I end up just being demotivated and that’s where I’m at right now. Felicedev (@felicedev) · 2025-08-16 · ♥ 4 Really, the situaton with Agent mode is so bad, it can’t solve a login with Supabase integration, i waste all 16$ credit + 20$ gifted credit and the situation still the same, redirect wrong at login and project ruined. Don’t more use V0 in my project, start in Free using 5$ with good and acceptables performance, when upgrade to Premium performance became really bad, and waste all the credits fast to resolve one single issue and sometimes lose all the project, now with Agent mode the quelity of work are really bad and also stuck in problem. Bohmebraden 1317 (@bohmebraden-1317) · 2025-08-16 · ♥ 4 Goodbye V0.. I will not give you guys a second chance. Agent mode is a failure. Sticking with Bolt/Base44/replit/lovable.. fine tune with cursor/copilot. V0 is trash. Mtspoelstra (@mtspoelstra) · 2025-08-16 · ♥ 4 Can you please restore the old version of v0 before agent mode? I am practically begging you at this point. Or at least communicate with your community and tell us if that is a possibility or not so we can move on and find other platforms to use. I have been trying to work with agent mode since you launched it and I am going around in circles - as soon as i get it to successfully add one feature (which takes WAY too long as it is), it breaks something else. And these are simple requests,. Not only have I accomplished NOTHING on my project since agent mode was launched, I actually went backwards….there is so much broken now I don’t even know where to begin fixing things…and I don’t seem to be able to successfully restore old versions. I have wasted a week trying to get this to work for me and I am desperately hoping you will please please please give us an option to disable agent mode. You used to be the best platform for AI coding and now you are literally unusable. Karim (@jdkmarketing) · 2025-08-17 lol the exactly same happened for me with supabase, nice to know lipuu (@lipumiya14-1174) · 2025-08-17 it causing problem when i am trying to fix a simple error, its trying to rewrite whole new app when i just want it to fix my current script only edit:At least give us a quick toggle between Agentic mode and normal mode. I just wanted to fix a single file instead, it spun up a whole new webapp and drained all my credit in a loop. This kind of overreach makes it unusable for simple tasks. Please let us opt out or switch modes easily. mj (@milindaj-3376) · 2025-08-17 · ♥ 1 Agree, 2 of my projects got botched once the agantic mode was started. We should have an option to de-select this mode and also select preferred model and model size Team1grin (@team1grin-7008) · 2025-08-17 · ♥ 4 The agent mode sucks :sob: please revert back to the option to use GPT or other models. I’ve been on vacation the past week and just came back without noticing the change - the agent burned through $10 in credits in 2 generations and did none of the things I asked for. I have been a huge van of vercel, but this is going to make me switch :worried: 10x the price for less effective work is not worth it sdharma (@sohan-planpodcom) · 2025-08-18 · ♥ 2 Yes really unhappy with the latest update. I need to disable agent mode to continue using v0 Tzuyi 1494 (@tzuyiyu-1494) · 2025-08-18 · ♥ 2 I sincerely suggest that v0 team **at least provide the option to use the old model.** The fact is, v0 Agent Mode clearly doesn’t seem ready yet, which has caused many client projects to come to a halt. What’s worse, **we are paying for Agent Mode to handle problems that it simply cannot solve.** The v0 team should properly test the new model before rolling it out, instead of updating it directly and making customers test it for you—while not even offering the choice of the old model. This creates tremendous inconvenience. Extralsc (@extralsc) · 2025-08-18 This update is absolutely unacceptable! The quality of the code is horrendous, and the results are far below what I expected. It feels like a huge step backwards. Not to mention, the inability to test code on a new URL without an iframe modal? Seriously? There needs to be an option to revert this update ASAP. I’ve had enough, and I’ve already canceled my subscription. This is a major disappointment, and I’m seriously hoping the developers recognize how badly this has missed the mark. KIDSHAKER (@kidshaker) · 2025-08-18 · ♥ 5 # The New V0 Agent is a Complete Disaster - My Personal Experience I'm absolutely frustrated and need to share this. The new Agent feature is a catastrophic regression that has completely ruined V0 for me. ## What's Broken **Everything I built before is screwed**: All my previous projects can't evolve properly anymore. The Agent makes terrible architectural errors and gives recommendations that make no sense. It mixes technical and functional aspects in ways that show it doesn't understand what it's doing. **Vibe coding is dead**: What made V0 special - that smooth development experience - is completely gone. It's like they took everything that worked and broke it. ## I've Already Moved to Bolt Switched to Bolt.new and there's no comparison. It just works. My team and I wasted hundreds of hours on V0 projects that are now basically unusable. ## Where's the V0 Team? The technical team's lack of response is unacceptable. We loved using V0, but this silence while everything falls apart is just unprofessional. **John Smith (CTO) - what are you doing about this?** At minimum, we deserve an explanation and feedback about what went wrong. ## I'm Done I'm canceling my subscription and requesting token refunds. This is a complete failure on V0's part. It's disappointing because I really wanted this tool to succeed. Anyone else dealing with this mess? How are you handling the migration away from V0? --- *Long-time V0 user, now former subscriber* Gwong492 (@gwong492) · 2025-08-18 I previously used the GPT-5 model and found its code to be bug-ridden, often requiring manual fixes. When I tried prompting it to correct its own errors, it underperformed, consuming tokens without resolving issues. At the time, switching to the v0-1.5-md model was an easy solution, as it produced reliable code for my needs, such as building a fairly in-depth app (still in development but functioning as designed) for just 25 credits. Now, the Agentic mode, which appears to rely on GPT-5, is producing similar buggy code that it struggles to fix. For example, I spent 50 credits on this app that still doesn’t work as intended, a stark contrast to my experience with v0-1.5-md. The forced shift to Agentic mode, without the option to select a model, has left me in a tough spot. The higher token costs and error-prone outputs make it hard to justify continuing with v0. I’m now weighing whether to keep using v0 in hopes that Agentic mode improves or to move my codebase to a platform that offers more predictable token usage and model flexibility. Has anyone found workarounds for optimizing Agentic mode, or can Vercel consider restoring manual model selection and per-chat token tracking to address these issues? Will Vercel reimburse credits for consistent errors caused by the reliance on the Agentic model? Justasjasaitis (@justasjasaitis) · 2025-08-18 · ♥ 3 Hi, Can you plesae make option to turn off that Agent mode, as design quality is pretty bad now. LeadLegend (@support-5002) · 2025-08-19 · ♥ 1 I was playing around with the Agent mode to test it with the 20usd of free credits and all of a sudden they vanished, of course i was not notified so it just started consuming my bought credits. Am i going to be refunded for this? did this happen to anyone else? And does the team listen to anyones feedbac? Everyone hated the three tier model when we all signed up with premium plans for unlimited messaging but they didnt care. Now they put the agent mode which everyone also seems to hate and they force us to burn through credits like we are some VC backed vibe coders with money to burn. Do better guys Ash (@crossmedia) · 2025-08-19 Yes, I was excited when I was able to manually switch on GPT-5 but it was ponderous, buggy and I quickly lost confidence in it. It is as if GPT-5 is overthinking or something. I’ve got the same dilemma as you in that I really like the v0 platform and its integration with Vercel platform, but the Agent is driving me nuts. It is simply not as intelligent as the previous models. Even if they are secretly routing the code-writing to the old model, something or other is impeding quality and speed. It feels like a backward step. Erikvandergeld (@erikvandergeld) · 2025-08-25  Agent mode keeps reviewing the same code, again and again, unfortunately. Meiko (@romanwebz) · 2025-08-26 Ok, so after long work, the update made it worst. I don’t know why, its giving more issues and more loosing money from my side. so…. wish for better updates later ) Vercel team you can look on my least chat to understand. Wish for updates and may be ill back ;) Andybeast (@andybeast) · 2025-08-26 If I’m being honest, I find the new Agent mode underwhelming compared to the old V0 setup. The previous version felt more focused, and I didn’t end up spending 50 cents on a prompt that wasn’t answered well. Unless there’s an option to return to the old Vercel mode, I will find another provider. CMA (@anton-4938) · 2025-08-27 · ♥ 3 I’ve opened a thread a few days ago about this, but there was no reply. So I’ll post here as well. I’ve been using V0 for quite a while and had great results and got my team to use it as well, but since introducing Agent, we’ve noticed a sharp increase in hallucinations as well as cases where requests are either misinterpreted or ignored repeatedly. When asked for very basic requests, or bug fixes the agent model decides to refactor (badly) the entire project as well as and redesign it from scratch. This results in the need to keep reverting back to previous versions time and time again. I’ve found myself (and my team members) spending significantly more time correcting these kinds of errors and misinterpretations, all while burning through tokens because the model is inefficient and ineffective. It randomly stops, goes into loops, makes completely random changes and more, do other members have the same experience here? Model releases can be difficult and some optimization time is granted, but the level of hallucinations and bad output makes it quite unreliable for quick iterations and ongoing work. For example, I asked for a color palette update, the component library was switched, the app was refactored and the design was changed. None of this was necessary or something that I wanted. Furthermore when I reverted back, and asked a simple question to verify the library the was used (to give it context once again), the model went into a loop and decided to refactor and change the component library. In the previous version, these kinds of issues were far less frequent and the output was much better overall. Please advise, Thanks! Sidsahgal 1261 (@sidsahgal-1261) · 2025-09-05 · ♥ 1 This agent mode is awful and so so slow. v0 used to work so well - please go back to the earlier version. Will have to move to another tool. Alexcumbers 9492 (@alexcumbers-9492) · 2025-09-05 Exactly same here, it’s a disaster when it was awesome before, we can only hope the Dev’s are seriously figuring this out, it looks like their response will come Monday: https://community.vercel.com/t/community-session-v0-product-announcement/21743
