JetBrains AI
Supercharge your tools with AI-powered features inside many JetBrains products
FAQ: New AI Quotas
We hear you. And you’re right – the new quotas are a reduction from what you experienced. When we launched, we had little usage data for agentic experiences and no top‑up feature existed, so we set quotas much higher than the real cost of AI usage (tokens from providers) as a temporary measure. That felt great, but it loses money at scale and puts the product at risk. We set expectations incorrectly. That’s on us.
Check more details about the new quotas in the original blog post.
Your Most Frequent Questions
“It’s not transparency; it’s paying more for less.”
First, cost. There’s no scam here. We’re aligning usage to real, public provider prices per token. Whether you run through us or directly with providers, it will be similar. Discounts vary, and yes, some companies still burn VC money to attract users (also that’s changing). We’re a real business; we can’t play that game. This is the real price of AI. Our job now is to make sure the value you get from our product far exceeds what you spend.
Second, transparency. We’ll roll out UX updates with statistics in a few weeks, and more usage details and controls after that. Being transparent about a bad situation is better than having unclear or uncontrolled terms. Our previous quotas were a temporary measure, and so they were not transparent. With the new approach, the rules of the game are clear. This is a sustainable foundation that will help us look further into the future and design new products and services around it.
“I can’t estimate how many credits I need for specific tasks.”
AI usage isn’t a flat fee because it depends on the type of model you use and the amount of text (tokens) flowing in and out. That means the same question can cost differently depending on how it’s phrased, how much context is added, or how long the conversation goes. In agentic workflows, where the AI may iterate and correct itself, usage can naturally be higher.
That said, you don’t have to fly blind. We aim to give you the tools to get the best grasp on AI usage and make sure you stay in control:
- Clear pricing per credit: We align with public provider costs, so you always know the unit price.
- Transparent usage stats: We’re rolling out UX updates that let you see exactly where credits go. It will roll out in waves, with the first wave in the next few weeks.
- Best-practice guides: We will share ways to phrase prompts and structure workflows so you get quality results faster (and spend less).
- Safeguards: Subscription quotas and alerts help prevent unexpected overuse, and you can always top up to continue your work.
Think of credits like fuel for your AI workflows: you may not know the exact mileage before the trip, but you’ll always see your fuel gauge and can plan accordingly.
“My quota runs out in half a day. Is it a bug?”
No – the quota got smaller for both AI Pro and AI Ultimate. In our data, about 80% of users (plan and usage dependent) won’t hit the new limits. The downside is that the most active AI users are the most affected.
Again, we’re not targeting you and we’re not slicing by some percentile; we’re aligning with costs so we don’t burn money indefinitely, while still giving away more (e.g. AI Ultimate includes roughly 15% more value than you pay each month, and annual plans can go up to about 40% more than your pay).
There’s another reason quotas can vanish fast: agentic workflows. Agents collect context and iterate through a plan; that means multiple, long requests that eat tokens. Junie, our coding agent, is built for quality on real‑world dev tasks. Speed and cost are part of the product. Early on, we took this risk on us. That was fine for kick‑off – but it’s not how this works long‑term.
Junie is designed for complex tasks and provides high quality output. Code generation, verification and tests together take more AI credits – but it also helps Junie achieve the better result for your job. We don’t want to sacrifice quality over speed: we benchmark the models and choose the best for your tasks, we always prioritize Junie’s ability to compare the solutions and offer you the most suitable one. We train Junie to verify the steps it is taking – not just to generate code faster, but to create solutions that you can use and maintain. All this helps Junie perform like agents in more expensive plans (up to $200/month), but we aim to keep the cost of usage reasonable and constantly work on optimization.
Smart ways to think about your AI quota
- Track your usage (detailed stats are coming soon; for now, watch your in‑app meter).
- Consider the best tool for the task – whether it’s an AI agent (complex workflows, iterations, validation, etc.) or an AI assistant (more straightforward Q&A and one-step solutions).
- Be intentional with the way you prompt and the model choice (e.g., GPT‑5 is often cheaper than Claude 4 with comparable results – try switching if it works for you).
- Start over instead of a follow-up if you feel that the agent is lost. Based on the way LLMs work, multiple follow-ups (e.g., more than seven) usually drift away and take more credits (the longer the chat, the more context is passed with every request) but may not get you closer to the result you have in mind.
- Favor smaller/cheaper models for drafts (check the hints in AI Chat), then upgrade for final runs.
“I just paid for an annual AI subscription based on the old patterns. Why no refund?”
We do have a refund policy, and some of you have already received refunds that extended your non‑AI subscriptions. Given this, we understand the frustration. Even if the refund window has passed, we’ve prepared a set of special alternatives for annual AI Pro and AI Ultimate users; you can select the one that works best for you. Keep in mind that we can’t grant refunds for All Products Pack and dotUltimate beyond regular policies, because AI Pro was included there at no extra charge.
That said, we still think annual subscriptions are a good deal. Example: AI Ultimate (annual) gives you 420 AI Credits for $300 — roughly 40% more credits than your payment at public rates.
***
AI is a great tool, and it’s changing how we build software. But it’s not magic — especially not free, unlimited magic. As professionals, we should understand how it works, how much it costs, and how to use it better. In the real world, speed and cost matter as much as quality. We believe that’s where the industry is heading. So let’s face it and start mastering these together.
Thank you to everyone who’s supported us – and for all the honest feedback. We’re building a sustainable business. Our future is tied to the value we create for you, so we’ll keep delivering exactly that step by step.
Subscribe to JetBrains AI Blog updates
