API Need
GGemini 3 Pro Image Preview API
gemini-3-pro-image-preview
Gemini 3 Pro Image Preview is Google’s preview-era Nano Banana Pro model for reasoning-driven image generation, visual editing, and mixed text-and-image responses.
Gemini 3 Pro Image Preview targets demanding creative work such as product mockups, diagrams, localized graphics, reference-guided compositions, and iterative visual refinement. The supplied APINEED contract exposes Gemini 3 Pro Image Preview through text-generation profiles and an asynchronous text-to-image workflow. Google later promoted the model line to the stable Gemini 3 Pro Image identifier, so teams maintaining Gemini 3 Pro Image Preview integrations should evaluate outputs carefully and keep migration tests ready.
MODALITIES
Playground
Generate with gemini-3-pro-image-preview, review the output, and use the same request shape in your application.
No supported image request format is published for this model yet.
Output Ready
Request preview
{
"model": "gemini-3-pro-image-preview",
"workflow": "text_to_image",
"input": {
"prompt": "A refined product photograph with natural lighting, realistic materials and a clean background."
},
"output": {
"count": 1
}
}Provider
Current API Need routing.
Pricing
Current selling prices in USD. Channel and group adjustments are already included. A range means available routes have different prices.
Image output$78 – $125.97 / 1M image tokens
Input$1.3 – $2.0995 / 1M
Text output$7.8 – $12.597 / 1M
Cache read$0.13 – $0.20995 / 1M
Cache write$1.54375 – $2.49315625 / 1M
Web search$9.1 – $14.6965 per thousand queries
Availability
Recent public performance data for this model.
No model-specific availability history is published yet. A successful request is not an uptime guarantee.
Gallery
Quick Start
Call gemini-3-pro-image-preview with the endpoint and request format shown below.
Create an API key
Store your APINEED key server-side as API_NEED_API_KEY.
export API_NEED_API_KEY=sk-apineed-v1-...Send the request
Use gemini-3-pro-image-preview with POST /v1/media/generations.
curl https://apineed.com/v1/media/generations \
-H "Authorization: Bearer $API_NEED_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gemini-3-pro-image-preview",
"workflow": "text_to_image",
"input": {
"prompt": "A refined product photograph with natural lighting, realistic materials and a clean background."
},
"output": {
"count": 1
}
}'Poll the task
Save id, then poll GET /v1/media/generations/{id} every few seconds until the task is succeeded or failed.
curl https://apineed.com/v1/media/generations/$TASK_ID -H "Authorization: Bearer $API_NEED_API_KEY"const images = task.outputs.filter((item) => item.type === "image").map((item) => item.url);gemini-3-pro-image-preview API Endpoint
Create an asynchronous image task, then use its task ID to retrieve status and direct object-storage output URLs.
POST
/v1/media/generations- Authorization
- Bearer $API_NEED_API_KEY
- Content-Type
- application/json
- HTTP-Referer
- optional - your site URL, for rankings
- Model
- gemini-3-pro-image-preview
GET
/v1/media/generations/{id}- Authorization
- Bearer $API_NEED_API_KEY
- Content-Type
- application/json
- HTTP-Referer
- optional - your site URL, for rankings
- Model
- gemini-3-pro-image-preview
Parameters
Use the supported parameters below. Leave optional fields out to use the service defaults.
| Name | Type | Values | Description |
|---|---|---|---|
model | string | Required | Use gemini-3-pro-image-preview. |
workflow | string | text_generation, text_to_image, image_edit | Use an APINEED-supported workflow. |
input.prompt | string | Required | Describe the requested image. |
input.assets | array | Required for image_edit | Image URLs with role input_image; up to 1 references. |
Gemini 3 Pro Image Preview API Introduction
A researched guide to Gemini 3 Pro Image Preview, its capabilities, integration fit, and operating considerations.
What is Gemini 3 Pro Image Preview?
Gemini 3 Pro Image Preview is the preview model identifier originally released for Nano Banana Pro, Google’s higher-fidelity image generation and editing model built on Gemini 3 Pro. Google introduced Gemini 3 Pro Image Preview for complex visual production rather than high-volume commodity generation. Its strengths include interpreting detailed instructions, combining language reasoning with visual synthesis, placing requested text inside designs, and refining an image through conversation. Google describes the underlying model as suitable for graphic design, product mockups, factual visualizations, diagrams, and other assets requiring precise creative direction. Gemini 3 Pro Image Preview should nevertheless be treated as a generative system rather than a deterministic renderer: a strong prompt can improve compliance, but it cannot guarantee an exact layout, factual statement, spelling, likeness, or brand treatment on every run. [2][1][7]
Inputs, outputs, and context limits
Gemini 3 Pro Image Preview accepts text and image inputs and can return text and images. This enables both direct text-to-image requests and workflows in which an uploaded reference supplies composition, subject, branding, texture, or editing context. Google documents a 65,536-token input limit and a 32,768-token output limit for the stable Gemini 3 Pro Image successor, which preserves the Nano Banana Pro model line represented here. Treat these limits as upper bounds, not ideal prompt sizes: shorter, structured briefs are usually easier to inspect and reproduce. Gemini 3 Pro Image Preview works best when the request separates required content from optional style choices and identifies which supplied image is the primary subject, reference, mask, or layout guide. Output may interleave explanatory text with generated visual material where the selected workflow supports it. [1][6][4]
Where Gemini 3 Pro Image Preview fits best
Use Gemini 3 Pro Image Preview when visual quality and instruction depth matter more than the fastest possible response. Strong candidates include advertising concepts, packaging studies, editorial illustrations, presentation graphics, storyboards, product visualization, multilingual posters, interface mockups, diagrams, and educational infographics. Gemini 3 Pro Image Preview can also combine multiple visual references into a unified composition, which is useful when preserving a product, logo, character, palette, or design language across drafts. Google positions the underlying Pro model for challenging, multi-turn creation and modification, while newer Flash image models emphasize efficiency and throughput. That distinction makes Gemini 3 Pro Image Preview more appropriate for deliberate asset-production pipelines than for latency-sensitive interfaces that generate large numbers of disposable variations. [2][1][4]
Prompting Gemini 3 Pro Image Preview
Prompts for Gemini 3 Pro Image Preview should describe a scene rather than provide an unstructured keyword list. State the subject, environment, composition, camera perspective, lighting, materials, color treatment, output purpose, and any text that must appear. Put literal copy in quotation marks, specify its location and typography, and ask for only the text that belongs in the image. For edits, say what must change and what must remain untouched. Gemini 3 Pro Image Preview benefits from progressive iteration: establish the composition first, then refine details, instead of rewriting the entire brief after every result. When reference images are included, assign each one a clear role. Even with detailed instructions, review spelling, small objects, hands, spatial relationships, and brand details before accepting an asset. [8][4][9]
Conversational editing and visual consistency
A major reason to select Gemini 3 Pro Image Preview is its support for conversational image refinement. Developers can preserve a draft, request a localized adjustment, inspect the result, and continue refining without restarting the concept. Google’s Gemini 3 guidance says thought signatures are important for multi-turn editing with Gemini 3 Pro Image Preview because they preserve information about the prior image’s composition and generation logic. Applications using a compatible conversational protocol should therefore retain the complete model response and associated metadata rather than reconstructing history from visible text alone. Gemini 3 Pro Image Preview can improve continuity, but identity, geometry, typography, and fine details may still drift over repeated edits. Save accepted checkpoints and compare each revision against explicit invariants. [5][4][7]
APINEED integration and execution behavior
The supplied APINEED contract identifies Gemini 3 Pro Image Preview with the executable model name gemini-3-pro-image-preview. Text generation is represented by synchronous and streaming profiles on /v1/messages and /v1/chat/completions, plus synchronous generation on /v1beta/models/{model}:generateContent. The advertised text-to-image profile uses POST /v1/media/generations asynchronously and polls /v1/media/generations/{id}. Applications using Gemini 3 Pro Image Preview should follow the chosen profile’s response lifecycle rather than assume all requests complete inline. The capability record also lists image editing, but this snapshot contains no separate executable image-edit profile, so integrations should not invent an editing route. Preserve the model identifier exactly, store asynchronous job IDs durably, apply bounded polling with backoff, and handle policy rejection or incomplete media as normal outcomes. [4][1]
Evaluation, reliability, and latency
Evaluate Gemini 3 Pro Image Preview with prompts drawn from the intended product, not only attractive demonstrations. A useful test set covers prompt adherence, subject consistency, text accuracy, localization, prohibited additions, reference fidelity, edit locality, factual correctness, and output diversity. Human review remains important for perceptual quality, while automated checks can validate dimensions, file decoding, required phrases, response structure, and obvious policy failures. Gemini 3 Pro Image Preview performs internal reasoning and high-resolution visual synthesis, so latency can vary with prompt complexity, reference count, requested detail, safety processing, and output size. Avoid promising a fixed completion time. Use request timeouts, retry only transient failures, prevent duplicate jobs with idempotency controls, and record the model name, prompt version, inputs, result status, and reviewer decision for reproducibility. [1][10][6]
Lifecycle, safety, and production governance
Google’s lifecycle table records Gemini 3 Pro Image Preview as a preview released on November 20, 2025, with Gemini 3 Pro Image designated as its stable replacement. This distinction matters for regression testing because a stable successor may preserve the product concept while changing visual style, prompt interpretation, latency, or safety behavior. Deployments retaining Gemini 3 Pro Image Preview should isolate the model identifier in configuration and maintain comparison fixtures for migration. Google also advises developers to perform application-specific safety analysis, adversarial testing, monitoring, and user-feedback review. Generated images include SynthID watermarking according to the image-generation documentation. Gemini 3 Pro Image Preview outputs should still pass through product-level moderation, provenance handling, rights review, and human approval when they affect people, brands, regulated decisions, or public factual communication. [3][4][10][11]
Gemini 3 Pro Image Preview API Q&A
Practical answers for developers evaluating and integrating Gemini 3 Pro Image Preview.
Which model ID should I use for gemini-3-pro-image-preview?
Use gemini-3-pro-image-preview, exactly as returned by the API Need pricing API.
Which endpoint should I use?
Send a POST request to /v1/media/generations using your APINEED API key. Save the returned task ID, then poll GET /v1/media/generations/{id} until the task succeeds or fails. Follow the request example and parameter table on this page.
How is the API Need price calculated?
See Pricing for the current rates and discounts. Final charges use recorded usage.
Can Gemini 3 Pro Image Preview generate and edit images?
Yes. Gemini 3 Pro Image Preview belongs to Google’s Nano Banana Pro image model line, which supports generation from text, generation using image references, and conversational editing. In this APINEED snapshot, text-to-image has an explicit asynchronous execution profile. Although image editing appears in the capability contract, no separate executable editing profile is supplied, so do not assume an undocumented APINEED path. [1][4]
Does Gemini 3 Pro Image Preview return text as well as images?
Gemini 3 Pro Image Preview supports both text and image outputs according to the model family documentation and supplied capability contract. The actual response shape depends on the APINEED execution profile selected by the application. Validate content types instead of assuming that every successful response contains an image, because a model response may include text, a refusal, status information, or mixed content. [1][6]
Is Gemini 3 Pro Image Preview good at rendering words inside images?
Google highlights improved text rendering as a central Nano Banana Pro capability, making Gemini 3 Pro Image Preview suitable for posters, diagrams, labels, mockups, and localized graphics. It is still important to verify every character. Keep copy concise, place exact wording in quotation marks, describe typography and placement, and use post-processing when legally or operationally exact text is required. [2][8][9]
How should I preserve context when editing with Gemini 3 Pro Image Preview?
Keep the full conversational response and relevant metadata between turns. Google specifically notes that thought signatures are critical to conversational editing with Gemini 3 Pro Image Preview because later changes rely on information retained from the earlier result. Store approved intermediate images as checkpoints too, since repeated edits can gradually alter details that the user intended to preserve. [5]
What workloads are a poor fit for Gemini 3 Pro Image Preview?
Gemini 3 Pro Image Preview is a weaker fit when ultra-low latency, deterministic pixel placement, guaranteed typography, or very high-volume inexpensive variation is the primary requirement. Conventional graphics tools remain preferable for exact templates and final typesetting. Google positions newer Flash image models as the more efficient option for throughput-oriented workloads, while the Pro line emphasizes complex production quality. [5][4][1]
How can I improve reference consistency with Gemini 3 Pro Image Preview?
Give each reference a named purpose, such as primary product, character identity, palette, logo, or composition guide. Tell Gemini 3 Pro Image Preview which attributes must remain fixed and which may change. Use clean references, avoid contradictory examples, and evaluate several seeded business scenarios. Reference guidance improves consistency but does not guarantee an exact identity, logo, or geometric match. [2][8]
Should factual infographics from Gemini 3 Pro Image Preview be trusted without review?
No. Gemini 3 Pro Image Preview can use reasoning and, in provider-supported contexts, Google Search grounding to create context-rich visualizations. Grounding can improve factuality but does not make every label, number, citation, or visual relationship correct. Verify facts against authoritative sources, proofread text, and require domain review for medical, legal, financial, safety-critical, or public-information graphics. [1][2][10]
What safety controls should surround Gemini 3 Pro Image Preview?
Combine provider protections with application-specific controls. For Gemini 3 Pro Image Preview, screen inputs, inspect refusals and safety metadata where exposed, moderate outputs, protect uploaded personal data, and test foreseeable abuse cases. Google recommends iterative safety testing and monitoring rather than relying only on built-in filters. Human approval is appropriate for sensitive likenesses, brands, political material, and high-impact decisions. [10][11][12]
How should applications handle Gemini 3 Pro Image Preview latency and failures?
Treat Gemini 3 Pro Image Preview generation as variable-duration work. For the APINEED media profile, persist the returned job identifier and poll the documented status path with bounded exponential backoff. Set an overall deadline, distinguish transient transport failures from policy or validation failures, and avoid blind retries that create duplicate images. Log enough metadata to diagnose failures without retaining unnecessary sensitive inputs. [6][10]
What should teams know about the Gemini 3 Pro Image Preview lifecycle?
Google records Gemini 3 Pro Image Preview as a preview release dated November 20, 2025 and identifies the stable Gemini 3 Pro Image model as its replacement. Keep the APINEED identifier gemini-3-pro-image-preview unchanged for this catalog integration, but separate it from prompt templates and business logic. Run side-by-side quality and safety evaluations before adopting any successor. [3][1][2]
How to deploy gemini-3-pro-image-preview
Connect gemini-3-pro-image-preview to your application through the APINEED media API.
Add credits
Add APINEED credits before submitting a billable generation.
Create an API key
Create and securely store an APINEED API key.
Configure the endpoint
Send gemini-3-pro-image-preview requests to /v1/media/generations.
Track the task
Poll the returned task ID and store the completed image URL from outputs[].