WARSAW, August 17, 2026, 07:14 CEST
- Google lists August 17 as the shutdown date for three stable Imagen 4 API endpoints.
- The recommended Gemini 3.1 Flash Image replacement is not a simple model-ID swap.
- A 1K replacement image costs $0.067, versus $0.02–$0.06 for Imagen 4.
Google, owned by Alphabet (NASDAQ:GOOGL), has reached the published shutdown date for three stable Imagen 4 endpoints. Developers still using them face failed production calls once Google completes the retirement.
The replacement changes more than a model name. Gemini 3.1 Flash Image uses a different generation interface, accepts more input types and charges by output resolution. For some workloads, the per-image price rises.
Google’s lifecycle page lists August 17 for the standard, Ultra and Fast endpoints. It calls Gemini 3.1 Flash Image the recommended replacement. The page describes shutdown dates as the earliest possible retirement dates, without publishing an hour.
| Retiring endpoint | Former role | Paid price per image | Published shutdown date | Recommended replacement |
|---|---|---|---|---|
imagen-4.0-fast-generate-001 | Fast generation | $0.02 | August 17, 2026 | gemini-3.1-flash-image |
imagen-4.0-generate-001 | Standard generation | $0.04 | August 17, 2026 | gemini-3.1-flash-image |
imagen-4.0-ultra-generate-001 | Ultra generation | $0.06 | August 17, 2026 | gemini-3.1-flash-image |
“Once a model is shutdown, it is completely turned off,” Google says. The endpoint then becomes unavailable. That makes fallback routing and error monitoring essential during the changeover.
Imagen 4 priced output by model tier. Gemini 3.1 Flash Image prices the image output by tokens, which scale with resolution. Google lists standard rates of $0.045 for 0.5K, $0.067 for 1K, $0.101 for 2K and $0.151 for 4K.
| Output option | Price per image | Cost for 10,000 images | Difference versus Imagen 4 Standard |
|---|---|---|---|
| Imagen 4 Fast | $0.020 | $200 | 50% lower |
| Imagen 4 Standard | $0.040 | $400 | Baseline |
| Imagen 4 Ultra | $0.060 | $600 | 50% higher |
| Gemini 3.1 Flash Image, 0.5K | $0.045 | $450 | 12.5% higher |
| Gemini 3.1 Flash Image, 1K | $0.067 | $670 | 67.5% higher |
| Gemini 3.1 Flash Image, 2K | $0.101 | $1,010 | 152.5% higher |
| Gemini 3.1 Flash Image, 4K | $0.151 | $1,510 | 277.5% higher |
The cost comparison is arithmetic based on current list prices. It is not a quality-equivalence claim. A Gemini request may include input-token charges, text output or paid search queries, depending on configuration.
The migration also expands what a pipeline can do. Imagen 4’s current model summary lists text input and image output. Gemini 3.1 Flash Image supports text, image and video context, conversational editing and outputs up to 4K.
| Pipeline element | Imagen 4 | Gemini 3.1 Flash Image | Migration impact |
|---|---|---|---|
| Primary API path | Specialized image-generation endpoint | generateContent or Interactions API | Request and response handling must change |
| Supported inputs | Text | Text, images and video context | New multimodal workflows become possible |
| Editing workflow | Not listed on the current model summary | Conversational generation and editing | State and reference-image tests are required |
| Output resolution | Not stated on the current model summary | 0.5K, 1K, 2K and 4K | Resolution now directly affects output cost |
| Search grounding | Not listed | Web and image search grounding supported | Search queries may add charges |
| Video-to-image | Not listed | Supported | Enables thumbnails and posters from video |
Google’s current image guide recommends the Nano Banana family for all new generation tasks. It positions Gemini 3.1 Flash Image as the general-purpose choice, balancing speed, quality and cost.
Teams should run representative prompts before shifting traffic. Tests should cover text rendering, brand controls, aspect ratios, safety blocks, latency and image consistency. A price check should use the actual resolution mix.
Vertex AI customers face a separate lifecycle. Google Cloud had already recommended moving its Imagen 4 endpoints before June 30, so Monday’s deadline concerns the Gemini Developer API schedule reviewed here.
Risks: Google does not publish the exact cutoff hour. A delayed retirement would not remove the migration need, while an immediate switch could expose untested code, cost spikes or output changes.
The practical deadline is now. Developers should remove all three Imagen 4 identifiers, validate the new API contract and monitor costs before restoring full traffic.
Further analysis
Which Imagen 4 API models are affected?
imagen-4.0-fast-generate-001, imagen-4.0-generate-001 and imagen-4.0-ultra-generate-001. Their published shutdown date is August 17, 2026.Will the endpoints stop at a specific hour?
What replacement does Google recommend?
gemini-3.1-flash-image, also called Nano Banana 2. It is the company’s general-purpose image model for new Gemini API workloads.Is migration only a model-name change?
generateContent or the Interactions API. Request structures, response handling, error paths and output checks need fresh testing.