Skip to main content
Agnes AI access plans, API keys, RPM limits, and subscription quotas Effective date: 2026-06-22
Video RPM update date: 2026-06-28
Version: Current public reference version
This document explains the current RPM limits for Agnes AI text models, image models, and video models, as well as Token Plan subscription quotas.

1. Access Types

Agnes AI currently supports three access types: A single user may have multiple access types at the same time. For example, a user may have:
  • Free / default API keys
  • Enterprise verified API keys
  • Token Plan API keys
Each API key type uses a separate limit pool.

2. Token Plan Quotas

In addition to RPM limits, Token Plan users are also subject to subscription quotas. RPM limits and subscription quotas apply at the same time.

Notes

  • Text model quotas are counted by request count.
  • Image model quotas are counted by generated image count.
  • Video model quotas are counted by generated video duration in seconds.
  • Token Plan users are subject to both RPM limits and subscription quotas.

3. Text Model RPM Limits

Notes

  • Free / default users are suitable for basic testing and low-frequency usage.
  • Enterprise verified users receive higher baseline text model access.
  • Token Plan users are suitable for production usage, high-frequency requests, and agent workflows.

4. Image Model RPM Limits

Image model RPM limits are based on resolution. In general, higher resolutions have lower RPM limits.

Notes

  • Image model RPM limits vary by resolution.
  • Higher resolutions consume more computing resources, so RPM limits are usually lower.
  • Token Plan users receive higher image generation limits for 1K / 2K resolutions.
  • Current 3K / 4K limits are stricter across all user types.
  • For image generation, 1K, 2K, 3K, and 4K are output tiers used with a supported ratio. They are not the same as common display labels such as 1080p, 1440p, or UHD.

Image Size Tiers

For predictable output sizes, request image generation with a tier-based size and a supported ratio. For example:
The example above uses the 2K RPM tier and returns 2624x1472. If you request unsupported exact dimensions such as 1920x1080 or 2560x1440, the service may normalize the request to the nearest supported tier and aspect ratio. For example, a 16:9 request can be normalized to the 1K output size 1312x736. For production workflows that need exact display canvases such as 1920x1080 or 2560x1440, use size: "2K" with ratio: "16:9" and then crop or resize the generated image downstream.

5. Video Model RPM Limits

Video models are subject to both RPM limits and daily video duration quotas.

Notes

  • Video generation has higher resource costs, so RPM limits are lower than text and low-resolution image models.
  • Token Plan users currently receive a video subscription quota of 500 seconds per day.
  • Video models are subject to both RPM limits and daily duration quotas.

6. RPM Limits vs. Subscription Quotas

RPM Limits

RPM means Requests Per Minute. It controls how many requests can be processed within one minute. Examples:
  • Free / default users have 1 RPM for video models.
  • Enterprise verified users have 2 RPM for video models.
  • Token Plan users have 5 RPM for video models.

Subscription Quotas

Subscription quotas control the total available usage within a specific time window. Examples:
  • Starter users can use 1,500 agnes-2.5-flash requests per 5 hours.
  • Token Plan users can generate 4,000 images per day.
  • Token Plan users can generate 500 seconds of video per day.

Both Limits Apply

Users must comply with both:
  1. RPM limits
  2. Subscription quota limits
For example, a Token Plan Pro user may have 1000 RPM for text models, but is still limited by the quota of 30,000 requests per 5 hours and 300,000 requests per week.

7. API Key Types

Different API key types can be used independently.

8. Limit Pools

Limits are shared by API key type, not by each individual API key. Creating multiple keys of the same type does not increase the total RPM or total quota.

Examples



10. FAQ

Q1: Can creating multiple API keys increase RPM or quota?

No. Multiple keys of the same type share the same limit pool. Creating more keys does not increase the total available RPM or quota.

Q2: Can Token Plan users still use free keys?

Yes. Token Plan users can still use free / default keys. Free keys and Token Plan keys use separate limit pools.

Q3: Can Token Plan users also become enterprise verified users?

Yes. After enterprise verification, users can use:
  • Free / default keys
  • Enterprise verified keys
  • Token Plan keys
Each API key type follows its own limit pool.

Q4: Do enterprise verified users automatically receive Token Plan quotas?

No. Enterprise verification only provides higher baseline RPM limits. Token Plan quotas are only available after subscribing to a Token Plan.

Q5: Is the image quota counted by request or by generated image?

The image quota is currently counted by generated image count. Token Plan users can currently generate:
  • agnes-image-2.1-flash
  • 4,000 images per day

Q6: Is the video quota counted by request or by video duration?

The video quota is currently counted by generated video duration. Token Plan users can currently generate:
  • agnes-video-v2.0
  • 500 seconds of video per day
Video RPM limits also apply.

Q7: Do different API key types share limits?

No. Limits are only shared among API keys of the same type. For example:
  • Free keys use the free / default RPM pool.
  • Enterprise verified keys use the enterprise verified RPM pool.
  • Token Plan keys use the Token Plan RPM and subscription quota pool.
Different API key types do not share limits.

11. Summary

Agnes AI currently supports three main access types:
  • Free / default users
  • Enterprise verified users
  • Token Plan users
They have different RPM limits for text models, image models, and video models. Token Plan users also receive additional subscription quotas, including:
  • Text request quotas
  • Image generation quotas
  • Video duration quotas
Multiple keys of the same type share the same limit pool. Different API key types use separate limit pools. All values in this document are current public reference values and may be adjusted in the future.