Urgent: Unexpected AI Credit Usage Issue

I'm seeing the same thing, when the app first free credits, it used a maximum of 12 credits per consultation.

After applying a full professional plan with 3000 credits, each consultation can now cost up to 200 credits ?¿?¿

i dont understand how it can be so disproportionate, some consultations have been completely wrong and required editing also… is this a marketing trick to burn out their subscribers later?

I don't get it; it seems a bit abusive and deceptive.

Hi everyone, Celine from the Figma Community here. Thanks for taking the time to share this, it's really helpful context.

Figma Make uses an agentic AI system, which means credit usage can vary based on factors like model selection, task complexity, and the amount of context provided.

On the higher credit usage you're seeing (like the 136 or 200 credits on a single prompt), a few key factors can impact this:
- Model selection: It's important to note that credit consumption may vary based on the model being used while prompting. More advanced models (like Claude Opus 4.60) are more powerful but they also consume more credits than the default model.
- Context materials: Providing additional context, such as attached files or design libraries, can help improve results but may also increase credit usage.
- Task complexity: More complex requests or prompts that require multiple steps will use more credits. Even shorter prompts can use more credits if the task involves broader changes across a large or intricate file.

Please also note that you can select the model for each prompt in Figma Make, giving you flexibility to manage credit usage based on your needs.

​@SKOUZ : Regarding the behavior you mentioned (e.g. fixes not applied or unrelated changes), that's definitely something we'd want to take a closer look at — especially if it's impacting both results and credit usage.
I saw that you have already reached out to the Support team, thanks for doing that! If you're able to share a specific example (prompt + result), along with the model used directly in your case, that would really help us investigate further.

Appreciate you raising this, it's valuable feedback for improving the experience.

Hi ​@Celine_ ,

thank you for your reply and for taking the time to explain the factors that can influence credit usage, this is appreciated.

To clarify our specific case:

Model selection: We have always used the recommended default model.
Context materials: None were provided (no attachments, libraries, or external files).
Task complexity: Extremely low - a simple prompt to change some label colors to black.

Brief scenario overview (see screenshots provided)

1. An error was introduced by the Agent itself after my simple prompt to change label colors (cost: 224 AI credits).
2. The Agent reasoned: “I see the issue! I accidentally removed the opening curly brace from the onChange prop. Let me fix that.”
3. The Agent attempted a fix and claimed success - however, the issue was not resolved.
4. The Agent continued reasoning, modifying multiple files (7 files, then 4 files), repeatedly stating that everything was fixed - which guess what - again was not the case.
5. Eventually, the Agent produced a working state - although later we discovered the fix was still incomplete.

Key concern

The root problem was introduced by the Agent itself, not by the original prompt.

In other words:

• A simple visual change request triggered a structural code issue
• The Agent attempted multiple self-repairs
• Additional prompts were required to fully resolve the situation

Overall, the correction process alone consumed an additional 200–300 AI credits, despite originating from an Agent-introduced error.

Why this is problematic

Compared to other AI tools we currently use, this level of consumption is exceptionally high.

Under the current logic, our Pro Plan allocation of 3,000 credits effectively translates to only about 20–30 prompts, which seems disproportionate for typical usage scenarios.

Even when extrapolating to higher-tier plans:

• Organization Plan: ~25–35 prompts (≈100 credits each)
• Enterprise Plan: ~33–43 prompts

This makes Figma Make AI approximately 10–15× more expensive than comparable tools for similar tasks, which is difficult to justify for ongoing production use.

Request for clarification

Could you please clarify what might cause such a spike in credit consumption under these specific conditions – particularly when:

• The default model is used
• No additional context is supplied
• The task itself is simple
• 2–3 months ago, AI credit consumption was much more consistent and transparent. Prompts typically triggered alerts such as: "Used 7 AI Credits" "Used 12 AI Credits" "Used 15 AI Credits" Occasionally, "No AI Credits were used"

Is there any mechanism in place to prevent excessive credit usage when the Agent is attempting to repair issues it introduced?

We genuinely see strong potential in Figma Make AI and would like to continue using it productively. However, the current cost-to-output ratio in scenarios like this raises concerns for sustainable adoption, especially at team scale.

Thank you again for your support and for looking into this matter.

Best regards,
Anton Elshani
SKOUZ, Germany

I’m experiencing the exact same issue and it’s honestly quite frustrating. The credit consumption for Make / AI tasks feels completely disconnected from the actual complexity of the work.

Just today, I had over 600 credits left. I prompted an adjustment to an existing design - a task that might cost a maximum of  200-300 credits based on my history. Not only did the AI fail to implement a single change, but it also wiped out my entire balance. I am now at 0 credits with zero output to show for it.

Charging users full price for failed iterations or simple corrections of mistakes the AI did - is not sustainable. There needs to be a refund logic or a “quality check” before credits are permanently deducted.

Thanks both for sharing these details, I hear your concerns around credit usage, let me clarify how this works.

On credit consumption: AI credits are based on the work Figma Make performs in response to a prompt (including follow-up prompts used to refine results). So even when iterating to correct outputs, credits will still be consumed accordingly if the system is performing actions.
(An "output" includes any work completed by the model such as generating or updating code, or making changes to your file). 
Please also note, if a prompt fully fails and no work is performed or no output is generated, credits aren't deducted.

Also, given the nature of generative AI, usage can vary depending on the situation. Even with the default model, results may sometimes be inaccurate, incomplete, or not behave as expected, and outputs can differ based on factors like the prompt or file context. However, credits will still be consumed if work was performed.

That said, the behavior you're describing is something we'd definitely want to investigate more closely.
For ​@SKOUZ:  since you already have an open support ticket, could you continue the conversation there? The support team is best equipped to review your specific case in detail.

For others experiencing similar issues — if you believe no work was performed or no output was generated, please reach out to our support team with a specific example (prompt + result) so our team can take a closer look at it. 

You can do this via the Support Hub → click Start a chat → type “Report a bug”, and include your impacted file by sharing it with support-share@figma.com with “Can View” access. This allows us to reproduce and investigate the issue.

Thank you again for raising this.