Serious Concerns Regarding JetBrains AI Pricing & Credit Consumption with Junie and AI Assistant
JetBrains — are you listening?
After reviewing the ongoing discussions in the community regarding Junie and the AI Assistant’s extreme credit consumption rates, it’s clear that many users are frustrated and seeking answers. Unfortunately, these concerns continue to go largely unaddressed. I recently submitted a bug report from within WebStorm after my credits were depleted with very minimal usage, and based on the community feedback, this issue is anything but new.
I have been a long-time WebStorm user and became a Junie subscriber several months ago. I intentionally moved away from Gemini CLI and GitHub Copilot—canceling both subscriptions—to fully commit to Junie because it provided comparable capabilities directly in the IDE. I upgraded from the Pro plan to the Ultimate plan because it was advertised as the option designed for developers who rely on AI daily. As a small developer, I watch my expenses carefully, and this seemed like the right consolidation: one product, one environment, predictable cost.
Then JetBrains abruptly changed the credit system and the way credits are consumed.
Earlier this month, I loaded a project and used Junie to review and refactor a small section of code. Before starting, I noticed I had 28 out of 35 credits available—a new metric, since previously usage had been displayed as a percentage bar. I didn’t think much of it at the time. But after only a few hours of normal use spread over a couple of days, my credits had dropped to 10 out of 35. For the level of usage I’m accustomed to, this was shocking.
When I contacted support, I received a boilerplate response indicating that the new system is “fair.” Fair for whom? Under the new structure, I would need to spend well over $100 per month to continue using the product the way I always had. Previously, the Ultimate plan worked perfectly. Shortly after the introduction of ChatGPT 5 capabilities, credit consumption escalated dramatically—without any announcement, explanation, or transparency.
Worse, there is still no clear explanation of what a single credit actually provides. I’m told I have 35 credits, but what does that equate to in practical, real-world usage? There is no clarity. Meanwhile, competitors such as GitHub Copilot offer unlimited use in the IDE for $39/month—only $9 more than your Ultimate plan—with no fear of running out of credits.
Because of this, I canceled my JetBrains AI license (License ID: O331N8V4TJ). I cannot justify the cost under the new model, especially when competing tools offer significantly more value and predictability. This is disappointing, because I genuinely liked Junie, supported it, and preferred the integrated experience. But the new pricing and usage structure has left me—and many others—no choice but to switch.
This is not an isolated concern. Your customers have been voicing the same frustrations for weeks! People are canceling. Developers are leaving. The sentiment is loud and clear: the new system is not sustainable for professional use.
If JetBrains revisits this pricing and credit model and restores something practical and predictable, I’m open to returning. But right now, the current structure is simply unworkable.
Please listen to your customers.
Please sign in to leave a comment.
Thanks for taking the time to share your very thorough feedback, we're sorry to hear the recent changes (https://blog.jetbrains.com/ai/2025/08/) have left you dissapointed. In hindsight, we should have communicated these changes better.
Hopefully, this FAQ (https://blog.jetbrains.com/ai/2025/09/faq-new-ai-quota/) covering some of the most common questions after the changes can help eliminate at least some of the confusion.
A rough estimation of what 1 AI Credit now gives you and how quota usage is calculated can be found here (https://www.jetbrains.com/help/ai-assistant/licensing-and-subscriptions.html#quota-faq).
There are also ways to reduce AI quota consumption at your disposal:
• Tips to control Junie's usage & quota limits (https://youtrack.jetbrains.com/articles/SUPPORT-A-1981).
• Use custom local models in AI Assistant (https://www.jetbrains.com/help/ai-assistant/use-custom-models.html) (these are not affected by your quota).
We are also working towards including better usage statistics so that you can see where your credits are going and stay in control.
If you have other questions or concerns about your particular credit usage, you are always welcome to contact our support (https://intellij-support.jetbrains.com/hc/en-us/requests/new?ticket_form_id=66731) for more information and assistance.
Best regards,
Leonardo
I’m currently looking for another IntelliJ plugin that consumes fewer credits than Junie.
I’m using the Gemini 3 Flash model through Junie, and according to JetBrains it’s supposed to be the most cost-efficient option — but honestly, it still burns credits very quickly.
At this point, it really feels like JetBrains has found a new way to monetize its users.
We are sorry that Junie doesn't meet your expectations. Currently, Gemini 3 Flash is our most cost-efficient option. If you would like to improve your usage costs, we recommend checking our article about controlling Junie's quota usage, available here: https://youtrack.jetbrains.com/articles/SUPPORT-A-1981/How-to-control-Junies-quota-usage-to-avoid-high-token-consumption . Did it help?
Regards,
Junie Support Team
I agree with the preceding sentiment. There is no warning or any way to estimate how much will be spent.
I do understand your concern that there is no warning or any way to estimate how much will be spent. However, quota usage is calculated dynamically for each interaction based on several factors, primarily the number of tokens used — units of text from both your input and the AI’s response. Since token usage can vary depending on the length, complexity, and context of a request, the quota cost may differ from one interaction to another.
If you have any other concerns, please feel free to open a ticket with our support team. https://www.jetbrains.com/support/
I concur with the OP's sentiment, I'm a very low AI user, just on a free 3 credit pack from having a license. But it's ridiculous how quickly these get used, I merely ask a couple of questions or get AI to write a commit message for me and pretty much most of my credit is gone.
Unfortunately this push by companies across the board to monetise AI will render it completely unusable/inaccessible for people like me.
With the AI Free tier you get 3 AI Credits every 30 days, and this is your full monthly cloud quota for JetBrains AI (AI Assistant and Junie). Even a few simple actions (a couple of questions, generated commit messages, or Junie runs on larger context) can use up this small allowance, so it’s expected that the quota may be exhausted after just a few days of occasional use.
To reduce consumption without upgrading, you can start a new chat for each separate topic so old context isn’t resent each time, prefer more cost‑effective models (for example, Gemini Flash instead of premium models like Claude Sonnet), and use Junie sparingly, as it tends to be more resource‑intensive and shares the same quota.
More details are available here: Licensing, plans, AI Credits and how quota is calculated: https://www.jetbrains.com/help/ai-assistant/licensing-and-subscriptions.html#ai-quota What leads to high quota consumption and how to reduce it: https://youtrack.jetbrains.com/articles/SUPPORT-A-2448
I'd like to add my experience to this thread, as it echoes the concerns already raised.
I'm on the AI Ultimate plan (30$/month, 35 credits). I recently asked Claude Agent to perform a task that I could have done myself in about 3 minutes - nothing complex, just a small, well-scoped change. It consumed 2 credits, i.e. roughly 2$ of my monthly quota for a trivial task.
At that rate, a developer doing a handful of similarly small agent requests per day would burn through the entire monthly Ultimate quota in a matter of days. This makes the agent unusable for exactly the kind of small, everyday tasks where AI assistance should shine.
Like others here, I still don't have a clear answer to a basic question: what does a single credit actually correspond to in terms of context size, tokens, or model calls? Without that transparency, it's impossible to predict or budget usage, and it's hard to justify the subscription cost when a competing tool offers unlimited usage for a similar price.
I'd really appreciate some clarity from JetBrains on:
1. What drives credit consumption so heavily on simple agent tasks
2. Whether there are concrete recommendations (beyond "start a new chat") to reduce consumption without losing functionality
3. Whether a more predictable/flat pricing model is being considered
Thanks for reading.
Dear Mjassani,
Thanks for taking the time to share your experience and detailed feedback.
AI credit usage depends on several factors: the selected model, how much context (files, previous messages, MCP data) needs to be processed, how many internal requests the agent makes to complete a task, and the size of the responses. Because this varies from task to task, a single credit can’t be mapped to a fixed number of tokens or calls. For concrete examples of what you can typically do with one AI Credit, you can refer to the official “Usage per AI Credit” section in our documentation:
https://www.jetbrains.com/help/ai-assistant/licensing-and-subscriptions.html#usage-per-ai-credit
To help manage your quota more efficiently, you can:
• Switch to a lower-cost model (for example, Gemini Flash) when appropriate, or consider GPT-5, which is often cheaper than Claude 4 while giving comparable results.
• Avoid putting very large files or attachments into context, and keep in mind that using MCP servers and large context attachments consumes more quota.
• Create new chats instead of using a single chat for multiple unrelated topics, because the number of tokens consumed grows quickly with long threads.
• Set a message trimming threshold in AI Chat so older messages are dropped from context when they are no longer needed:
https://www.jetbrains.com/help/ai-assistant/chat-mode.html#set-message-trimming-threshold
• Remember that Junie and Claude-based agents generally consume more quota than standard AI Chat.
For more guidance on what can lead to higher quota consumption and how to mitigate it, these articles may also be useful:
• https://youtrack.jetbrains.com/articles/SUPPORT-A-2448/What-can-lead-to-the-high-quota-consumption
• https://youtrack.jetbrains.com/articles/SUPPORT-A-1650
We understand that predictability and transparency around AI credit usage are important. There is an open feature request specifically for a more detailed quota usage report and statistics for JetBrains AI Assistant and Junie, which you’re very welcome to upvote and follow for updates:
https://youtrack.jetbrains.com/issue/LLM-19959/Add-a-detailed-quota-usage-report-statistics-for-JetBrains-AI-Assistant-and-Junie
If you have any other concerns, please feel free to open a ticket with our support team. https://jb.gg/ai-assistant-support
Thank you,
AI Assistant Support Team
I purchased 15 credits, and it used them in two questions. Thats an absolutely outrageous price. Nearly $11 AUD per question. No code generation used, and fairly minimal code inspection. How is any of this fair?
I've been a loyal subscriber for nearly a decade. But this is highway robbery and has me seriously questioning this whole subscription.
Thanks for sharing your experience, and for being a long-time subscriber.
To answer the core question directly: a credit is not tied to a fixed number of prompts or questions. Since the August 2025 update, AI credits reflect the real per-token cost of each request. How fast credits are spent depends mainly on two things: the model you choose, and how much context is sent with each request (open files, project context, conversation history, MCP data). Agentic workflows like Junie and Claude-based agents make several internal model calls and carry a lot of context to complete even a small task, so they consume noticeably more than a single AI Chat message. That is why a task that feels trivial can still cost more than expected.
A few things that help keep consumption down:
1. Use a lighter model (for example Gemini Flash) for routine work, and reserve premium models like Claude for when you really need them. GPT-5 is often cheaper than Claude for comparable results.
2. Keep the attached context small. Avoid attaching large files or the whole project when it isn't needed.
3. Start a new chat for a new topic so a long conversation history isn't resent with every message.
4. For code completion, the local Full Line completion runs on-device and does not consume cloud credits.
You can find more detail here: What can lead to high quota consumption (https://youtrack.jetbrains.com/articles/SUPPORT-A-2448) and Why is the cloud quota consumed faster than before (https://youtrack.jetbrains.com/articles/SUPPORT-A-2539).
We also hear the request for better transparency and predictability. There are open feature requests for exactly this, and you are welcome to vote and follow them for updates: LLM-16701 (https://youtrack.jetbrains.com/issue/LLM-16701) (track quota usage in more detail, including per-request cost) and LLM-19959 (https://youtrack.jetbrains.com/issue/LLM-19959) (a detailed quota usage report and statistics).
If you believe your usage is genuinely disproportionate to your activity, we would be glad to look into the consumption recorded on your account. Since the forum is not the right place to share account details, please open a request through https://jb.gg/ai-assistant-support with your AI license ID and a couple of specific examples (the model used and roughly when), and the team can investigate.
Jetbrains probably charges token usage based on API costs, which is way higher than the model providers subscriptions. That’s the only explanation I can think of. That said, I'm long gone from Jetbrains, unfortunately.
JetBrains charges for AI usage based on "AI Credits," which are directly tied to their own API costs for accessing third-party LLMs (like OpenAI, Anthropic, etc.). Each AI Credit is roughly equivalent to $1 USD, and the number of credits you get per month depends on your subscription tier. The actual quota usage is calculated dynamically, primarily based on the number of tokens processed (input + output), and can vary a lot depending on the model used, the length of your prompts, and the context size. More expensive models and agentic features (like Junie) consume credits much faster than basic code completions or explanations.
This means that, compared to a direct subscription with a model provider (like ChatGPT Plus or Claude Pro), JetBrains' pricing can feel much higher for the same amount of usage, especially since their quotas are often more restrictive and less transparent. Many users have voiced frustration about how quickly quotas are depleted, especially after recent changes that reduced the included credits and increased the cost per action. JetBrains says this is to reflect their actual provider costs, but the end result is that their AI features can end up costing more than using the same models directly.
For any other concerns, feel free to open a ticket with our support team: https://jb.gg/ai-assistant-support