Imagen API pricing looks simple when you only read the per-image row. In a production image workflow, the real decision is broader: which Google image model is current, whether an Imagen row is deprecated, which route family your gateway exposes, how the dashboard records the request, and whether the selected row passes a small smoke test before customer traffic.
This guide was checked on June 17, 2026 Asia/Shanghai against official Google Gemini API pricing, Google Vertex AI generative AI pricing, the Gemini API image generation guide, a fresh Ahrefs MCP run, and a live Flatkey pricing-catalog snapshot. Treat every number below as a dated planning input, then verify the current provider page and the current Flatkey pricing row before launch.
Quick Answer: Imagen API Pricing Checks Before Routing
Before you route an Imagen-style image request through an OpenAI-compatible workflow, confirm these items:
- Current model ID: verify whether the route uses
imagen-4.0-fast-generate-001,imagen-4.0-generate-001,imagen-4.0-ultra-generate-001, or a newer Gemini image model instead. - Lifecycle status: Google's Gemini API pricing page checked for this article marks the Imagen 4 IDs as deprecated and scheduled for August 17, 2026 shutdown.
- Pricing unit: normalize per-image price, image size, feature type, edits, retries, and route-specific billing fields.
- Flatkey catalog row: confirm endpoint types, group, ratio fields, availability status, and dashboard label on the day you test.
- OpenAI-compatible path: do not assume a chat-style request works for image generation. Validate the image endpoint, request body, response shape, and output storage.
- Usage proof: run a non-production request and confirm the usage log shows model, route, status, owner, and billable unit.
- Fallback rule: define whether a failed Imagen route may move to Gemini image, GPT Image, or another image model, and what output differences are acceptable.
That is the practical version of Imagen API pricing: price row plus route proof, not price row alone.
Current Imagen API Pricing From Google
Google's Gemini API pricing table checked for this article lists Imagen 4 as a paid-tier, per-image model family. The same section also includes a deprecation warning for the Imagen 4 model IDs, so pricing and lifecycle must be read together.
| Google Row | Model ID | Official Price Checked | Lifecycle Note |
|---|---|---|---|
| Imagen 4 Fast | imagen-4.0-fast-generate-001 |
$0.02 per image | Deprecated; Google lists August 17, 2026 shutdown. |
| Imagen 4 Standard | imagen-4.0-generate-001 |
$0.04 per image | Deprecated; Google lists August 17, 2026 shutdown. |
| Imagen 4 Ultra | imagen-4.0-ultra-generate-001 |
$0.06 per image | Deprecated; Google lists August 17, 2026 shutdown. |
The shutdown date is the most important part of this Imagen API pricing check. If your launch, contract, or customer workflow extends beyond August 17, 2026, do not treat Imagen 4 as a long-lived default without a migration plan. Google's pricing page points readers toward Gemini 2.5 Flash Image to avoid disruption, and Google's current image generation guide now centers Gemini image models such as gemini-3.1-flash-image, gemini-3-pro-image, and gemini-2.5-flash-image.
Why Per-Image Pricing Still Needs A Worksheet
For plain text-to-image generation, Imagen API pricing is easy to write in a table. Production workflows add extra choices that can change the unit you should budget against.
| Check | Why It Matters | Evidence To Save |
|---|---|---|
| Generation type | Text-to-image, edit, upscale, product recontext, and virtual try-on can have different rows in Vertex AI pricing. | Provider feature name and price section URL. |
| Resolution and format | A per-image row does not automatically describe aspect ratio, output size, file format, or storage behavior. | Request settings and accepted response format. |
| Retry policy | Rejected, failed, edited, or regenerated outputs can create more than one request for one user action. | Retry count, final status, and user-visible result. |
| Gateway unit | A gateway catalog may use ratio fields or routed unit labels that are not the same as the provider's public table. | Flatkey row, route, group, price label, and usage log. |
Google Cloud's Vertex AI pricing page also lists separate image-related entries, including Imagen 4 text-to-image rows, image editing, upscaling, product recontext, and virtual try-on. Use those rows as a reminder that a single Imagen per-image number is not a universal price for every image operation.
OpenAI-Compatible Workflow Checks
An OpenAI-compatible workflow can reduce client migration work, but it does not remove image-specific checks. A request shaped for a chat completion is not the same as a request shaped for image generation. For Imagen API pricing, the route contract should say exactly which endpoint family receives the request and which usage ledger records it.
Flatkey's public pricing API snapshot checked for this article exposed endpoint metadata for image-generation at /v1/images/generations, gemini at /v1beta/models/{model}:generateContent, and openai at /v1/chat/completions. The Imagen rows in the same snapshot listed image-generation, gemini, and openai endpoint types. That is useful catalog evidence, but it is not enough to approve production traffic.
Before launch, run a request that proves all of these fields:
- The model string in code matches the current Flatkey catalog row.
- The selected endpoint family accepts the image request body you plan to use.
- The response shape contains the output format your application expects.
- The dashboard usage log records route, model, status, owner, and billable unit.
- A failure case is visible enough for support and finance to audit later.
The adjacent Gemini API OpenAI-compatible router guide covers the broader base-URL pattern. This Imagen article narrows the gate to image-model pricing, deprecation, and endpoint proof.
Flatkey Catalog Snapshot For Imagen Rows
Flatkey's public pricing catalog was checked on June 17, 2026 Asia/Shanghai. The response contained 638 model rows and top-level pricing version a42d372ccf0b5dd13ecf71203521f9d2. Image-related rows included Imagen 4, Gemini image, and other image-generation models.
| Flatkey Row Checked | Endpoint Types Seen | Pricing Fields Seen | Status In Snapshot | Action |
|---|---|---|---|---|
imagen-4.0-fast-generate-001 |
image-generation, gemini, openai |
model_ratio: 37.5, completion_ratio: 1 |
unknown_failure |
Do not use as production proof without a fresh dashboard check and test request. |
imagen-4.0-generate-001 |
image-generation, gemini, openai |
model_ratio: 37.5, completion_ratio: 1 |
unknown_failure |
Verify against the Google deprecation warning before any new rollout. |
imagen-4.0-ultra-generate-001 |
image-generation, gemini, openai |
model_ratio: 37.5, completion_ratio: 1 |
unknown_failure |
Keep out of production until the row, request path, and usage log pass testing. |
gemini-2.5-flash-image |
gemini, openai |
model_ratio: 0.15, completion_ratio: 100 |
available |
Candidate migration row, but still test request shape and pricing unit. |
gemini-3-pro-image-preview |
gemini, openai |
model_ratio: 1, completion_ratio: 60 |
available |
Confirm preview naming, feature requirements, and current dashboard availability. |
This table is not a permanent Flatkey support claim. It is dated catalog evidence for a routing checklist. The right Imagen API pricing workflow is to compare official Google pricing, open the current Flatkey row, run a small request, and reconcile the usage log before expanding traffic.
Routing Decision Matrix
Use this matrix when deciding whether Imagen belongs in an OpenAI-compatible image workflow today:
| Situation | Suggested Decision | Reason |
|---|---|---|
| New production feature launching after August 17, 2026 | Prefer a current Gemini image row or another active image model. | Google lists Imagen 4 shutdown on that date. |
| Short internal test before the shutdown date | Allow only after row, endpoint, and usage-log checks pass. | Short-term testing can be acceptable if the team owns the migration risk. |
Flatkey row shows unknown_failure |
Block production traffic and request a fresh dashboard or product confirmation. | Catalog presence does not prove a working route. |
| Workflow needs edits, upscales, or product imagery | Check the exact feature-specific provider row before estimating cost. | Text-to-image price is not automatically the price for every image feature. |
| Fallback to another image model is allowed | Define acceptable output differences and budget limits before automatic routing. | Image model fallback can change style, text rendering, aspect ratio, safety behavior, and cost. |
For a broader unit-normalization view, use the AI image generation API pricing comparison. That article compares GPT Image, Gemini image, and Imagen units; this one is the provider-specific Imagen route gate.
Smoke Test Before Production
A responsible Imagen smoke test should be small and auditable:
- Open the current Google pricing and deprecations references for the model ID.
- Open the current Flatkey pricing page and record the exact row label, endpoint types, group, and status.
- Create a non-production Flatkey key or use a staging key with a strict quota.
- Send one low-risk image request through the intended endpoint family.
- Save the request settings: model, prompt class, size, quality or feature type, and output format.
- Confirm the response shape, generated output, dashboard status, billable unit, and owner.
- Run one controlled failure or blocked-row test if possible, then document the fallback rule.
If the smoke test cannot prove those fields, keep the route in research backlog. For image generation, unclear billing or missing logs create support and finance risk quickly.
FAQ
How much does the Imagen API cost?
Google's Gemini API pricing page checked on June 17, 2026 lists Imagen 4 Fast at $0.02 per image, Imagen 4 Standard at $0.04 per image, and Imagen 4 Ultra at $0.06 per image. The same section marks those Imagen 4 IDs as deprecated with August 17, 2026 shutdown, so confirm the current page before using those prices.
Does Imagen API pricing work in an OpenAI-compatible workflow?
Only after route validation. An OpenAI-compatible workflow can standardize access, but image generation still needs the correct endpoint family, request body, response shape, model row, and usage log. Do not assume a chat-completion request is valid for an image-generation route.
Can Flatkey route Imagen API requests?
The June 17, 2026 Flatkey catalog snapshot listed Imagen 4 rows with image-generation, gemini, and openai endpoint types. Those Imagen rows showed unknown_failure, so this is dated catalog evidence only, not a production-readiness claim.
Is Imagen 4 safe for new production builds?
Use caution. Google's current pricing page warns that the Imagen 4 model IDs are deprecated and scheduled for August 17, 2026 shutdown. New long-lived production builds should evaluate current Gemini image rows or another active image model before choosing Imagen.
What should I check in Flatkey before production traffic?
Check the exact model row, endpoint type, group, price label, status, dashboard usage fields, owner, quota, and a real smoke-test result. Imagen API pricing should be approved only after the current provider price and Flatkey route evidence agree.
Final Takeaway
Imagen API pricing is a useful starting point, but it is not enough for an OpenAI-compatible production workflow. Verify the model lifecycle, normalize the image operation, confirm the Flatkey route, and save dashboard-backed proof before routing customer traffic.
To compare the current image rows across your stack, start with Flatkey pricing, then use the smoke-test checklist above to decide whether Imagen is suitable for staging, migration, or backlog only.
