Compare image generation models and API pricing
Write a prompt. Pay per image.
Find models for text-to-image generation, image editing, creative production, and visual workflows. Price comparisons are only made when models expose a comparable per-image rate.
60 image models are listed today. All of them use one OneInfer API key and pay-as-you-go wallet.
Choose and call a image model in 4 steps
The list ranks them by their price per image. Different billing units are never mixed, so the order is fair.
Lowest comparable price: Stability AI: SDXL Turbo
Currently the lowest-priced option in the provider price per generated image group. Different billing units are never mixed.
$0.00022 / 1024x1024 image
View pricing and providersAvailable image models
Comparable models are ordered by their current normalized price. Pricing content refreshed November 9, 2026.
60 models
| Model |
|---|
Stability AI: SDXL TurboLowest pricedeepinfra SDXL Turbo is a distilled SDXL 1.0 from Stability AI that makes an image in a few steps, for fast, low-cost generation.
|
 deepinfra FLUX.1 [schnell] is Black Forest Labs' fastest open FLUX.1 model, a 12B rectified flow transformer that makes an image from a prompt in a few steps.
|
 minimax MiniMax Image-01 is a closed-source text-to-image (and image-to-image) generation model from MiniMax, available via API. Built on MiniMax's expertise in prompt adherence from the Hailuo video series, it delivers cinematic-quality images with precise prompt fidelity, advanced lighting, and photorealistic human subjects with natural skin textures.
|
deepinfra FLUX.2 Klein 4B is Black Forest Labs' compact, fast image generation and editing model designed for low-latency inference and efficient local deployment. It uses a 4-billion-parameter flow-based architecture and supports text-to-image generation, image-to-image editing, and multi-reference visual composition. The model is suitable for rapid prototyping, interactive creative applications, local image generation, and resource-conscious deployment. The open-weight 4B variant is available under the Apache 2.0 license.
|
deepinfra P-Image is Pruna AI's real-time text-to-image model, with clear text rendering and close prompt following in about a second.
|
deepinfra FLUX.2 Klein 9B is Black Forest Labs' fast, high-quality image generation and editing model built on a 9-billion-parameter flow-based architecture. It combines strong prompt adherence, realistic visual generation, and multi-reference image editing with a step-distilled architecture requiring approximately four inference steps. It is optimized for sub-second image generation, real-time creative workflows, interactive editing, and efficient deployment while providing higher visual quality than smaller Klein variants.
|
blackforestlabs FLUX 3 Image Klein is the 5.6B-parameter open-weight model in Black Forest Labs' FLUX 3 image family, released in October 2026. It creates and edits images from a prompt and up to 10 reference images, from 768 px squares up to 4K, for a fraction of FLUX 3 Image's price.
|
deepinfra FLUX.1 [dev] is Black Forest Labs' open 12B text-to-image model, with detailed output and close prompt following.
|
deepinfra FLUX.2 [dev] is Black Forest Labs' open FLUX.2 model. It makes images from a prompt and edits them with up to four reference images.
|
deepinfra P-Image Edit is Pruna AI's fast image editor: it inserts, removes and restyles content in up to four reference images from an instruction.
|
 deepinfra Ming-Image 0.1 Design Layer splits a flattened design into five transparent, editable layers (text, subject, background), each returned as an image and billed as one.
|
 bytedance ByteDance Seedream 5.0 Flash generates a single image from text and up to 10 reference images. Supports image editing and 1K, 1.5K, and 2K output.
|
deepinfra Bria RMBG 2.0 removes an image's background, leaving the subject on transparency. Trained only on licensed data.
|
 novita Qwen-Image-Txt2Img is a 14B-parameter diffusion model specialized in high-quality text-to-image generation with enhanced prompt understanding and stylistic versatility. Part of the Qwen multimodal family, it features advanced composition control, style adaptation, and detail preservation across diverse visual domains.
|
 grok Grok Imagine Image is xAI's speed-focused image generation and editing model, powered by the Aurora engine. Designed for rapid iteration, brainstorming, and high-volume workflows, it supports both text-to-image and reference-guided editing (accepting up to 5 reference images) in a single endpoint. It natively supports multiple aspect ratios to generate consistent layouts for social media and concept art.
|
 bytedance Seedream 4.0 is ByteDance's image creation model that unifies text-to-image generation and editing in a single architecture. It features 4K resolution output, enhanced reasoning capabilities for physical and temporal constraints, and supports multimodal inputs including text and images. The model excels in precise editing, multi-image reference generation, and advanced text rendering for infographics and layouts
|
fal Qwen-Image-3 is a Qwen image generation model for creating images from natural-language prompts. It is intended for visual content generation with support for prompt-driven composition, text rendering, and a range of creative image-generation workflows. Detailed architecture, parameter count, supported resolutions, and official benchmark results for this specific model have not been publicly verified.
|
deepinfra FLUX.2 Pro is Black Forest Labs' production-grade image generation and editing model, offering a strong balance of visual quality, prompt adherence, speed, and reliability. It supports photorealistic text-to-image generation, image-to-image editing, multi-reference composition, accurate typography, and detailed control over subjects, lighting, color, and composition. Designed for commercial creative workflows, product photography, advertising, marketing, and professional visual content.
|
deepinfra Wan2.6 Text-to-Image from Alibaba's Wan team, with precise style control, realistic portraits and long-prompt understanding.
|
deepinfra Wan2.6 Image Edit edits and combines one to four images from an instruction, with control over style, framing and lighting.
|
deepinfra Wan2.7 Image makes and edits images: from a prompt alone, or from up to nine reference images, with strong text rendering.
|
deepinfra Removes an image's background, guided by a prompt saying what to keep or remove, at the original resolution.
|
bytedance ByteDance's latest multimodal image generation model featuring 'deeper thinking'. Supports visual reasoning for complex spatial logic, multi-reference fusion (up to 14 reference images), and real-time online retrieval for up-to-date visual trends.
|
 Imagen 3 is Google's highest-quality text-to-image foundation model, generating images with rich detail, enhanced lighting, fewer artifacts, and accurate typography.
|
deepinfra Nano Banana 2 Lite is Google's fast, low-cost Gemini image model. It makes a 1K image from a prompt, or edits a reference image.
|
bytedance Seedream 4.5 is an upgraded image generation model from ByteDance, offering cinematic aesthetics, stronger spatial understanding, and richer world knowledge. It features improved consistency, smarter instruction following, and professional typography rendering. It supports 4K resolution output, multi-image reference control, and delivers generation speeds of 2-3 seconds
|
grok Grok Imagine Image 2.0 is xAI's flagship autoregressive image generation and editing model powered by the Aurora engine. It introduces precise prompt following, sharper typography rendering, and advanced editing features like localized segmentation and Smart Resize. Supporting up to 3 reference images for consistent character and style workflows across multiple 1K or 2K aspect ratios, it is built for top-tier creative control.
|
deepinfra FLUX1.1 [pro] is Black Forest Labs' proprietary text-to-image model, known for prompt following, visual quality and varied output.
|
deepinfra Bria 3.2 is a 4B text-to-image model trained only on licensed data, for commercially safe images with good text rendering.
|
deepinfra Bria 3.2 Vector makes vector illustrations from a prompt. Trained only on licensed data.
|
deepinfra FIBO 2.0 is Bria's text-to-image model with control over lighting, composition and camera, at 1 or 4 megapixels. It takes an optional reference image. Trained only on licensed data.
|
deepinfra FIBO Edit edits an image from an instruction. Send a mask as the second image to limit the edit to an area.
|
deepinfra FIBO Edit 1.5 edits an image from an instruction, or combines up to four reference images into one.
|
deepinfra Blurs an image's background while keeping the subject sharp.
|
deepinfra Removes an image's main subject or foreground objects and fills in the background behind them.
|
deepinfra Improves an image: sharper detail, balanced colour and more clarity, at 1, 2 or 4 megapixels.
|
deepinfra Extends an image beyond its borders to the aspect ratio asked for, generating the new pixels.
|
deepinfra Erases what a mask marks (send the mask as the second image) and fills in the background.
|
deepinfra Adds or changes objects in the area a mask marks (the second image), following a prompt.
|
deepinfra Replaces an image's background with the one in a reference image (the second image).
|
bytedance Seedream 5.0 Pro is ByteDance Seed's professional multimodal image generation and editing model designed for complex visual creation, precise image manipulation, and production-grade design workflows. It combines advanced reasoning with strong image-text alignment, structural coherence, high-density text rendering, and photorealistic visual quality. The model supports text-to-image generation, image-conditioned creation, multi-image fusion, spatially guided editing, sketch-based generation, color and material replacement, and intelligent layer separation. It also provides native multilingual prompting and text rendering across more than ten widely used languages, making it suitable for infographics, advertising, UI and graphic design, photography, localized creative content, and professional image editing.
|
deepinfra Ming-Image 0.1 Design is an open 6B text-to-image model for design work: legible interface and poster text, graphics, illustration and photography.
|
blackforestlabs FLUX 3 Image is Black Forest Labs' next-generation image synthesis and editing capability built on the FLUX 3 multimodal foundation. It is designed for photorealistic and stylized image generation, complex prompt understanding, accurate multilingual typography, diverse artistic styles, and reference-guided image editing. The model builds on a unified multimodal architecture shared with FLUX 3 video and audio generation, with improved composition and text rendering compared with earlier FLUX generations. Image availability is subject to Black Forest Labs' staged release and early-access program.
|
grok Grok Imagine Image Quality is the high-fidelity tier of xAI's earlier image generation lineup. It sacrifices raw generation speed in favor of producing photorealistic, highly detailed visuals with strong prompt adherence, accurate lighting, and cleaner textures. It is commonly used for final-quality renders that do not require the latest 2.0 feature set.
|
blackforestlabs FLUX.2 Flex is Black Forest Labs' controllable image generation and editing model, optimized for typography, graphic design, and production workflows requiring fine-grained control over the generation process. It supports text-to-image generation, multi-reference editing, adjustable guidance strength, configurable inference steps, and prompt upsampling. The model is particularly suited to posters, advertisements, product packaging, branded graphics, and text-heavy visual compositions.
|
deepinfra Clarity AI's creative upscaler enlarges an image and adds detail and realism, guided by an optional prompt.
|
deepinfra Clarity AI's Crystal upscaler enlarges portraits, faces and product shots with precise, natural detail.
|
Gemini 3.1 Flash Image, also known as Nano Banana 2, is Google's high-efficiency native image generation and conversational editing model. It is optimized for low-latency, high-volume image workloads and supports high-quality image generation, editing, improved text rendering, multiple aspect ratios, search-informed generation, and resolutions up to 4K.
|
deepinfra FLUX.2 Max is Black Forest Labs' highest-quality model in the FLUX.2 family, designed for maximum image fidelity, complex prompt adherence, detailed visual composition, and professional image editing. It supports text-to-image generation, multi-reference image editing, photorealistic rendering, accurate typography, and advanced composition control. The model also supports grounding search to improve generation involving real-world information. It is intended for premium commercial imagery, advertising, design, and demanding creative production workflows.
|
Gemini 3 Pro Image, also known as Nano Banana Pro, is Google's professional image generation and editing model with a reasoning core. It is designed for studio-quality visuals, complex compositions and layouts, accurate text rendering, and high-resolution creative production.
|
deepinfra Clarity AI's Flux upscaler enlarges an image with FLUX, preserving faces.
|
 oneinfer MAI-Image-2.5 is Microsoft's image generation and editing model designed for high-quality text-to-image creation and precise image-to-image editing. It supports prompt-driven image generation as well as targeted edits that preserve unaffected regions of an input image. The model is designed to produce realistic imagery with consistent visual structure and is suited for concept visualization, creative content generation, production design, advertising, image editing, and other professional visual workflows. It supports multiple reference images for editing and produces PNG image outputs with up to 1,048,576 total pixels.
|
oneinfer MAI-Image-2.5-Pro is Microsoft's highest-quality image generation and editing model in the MAI-Image-2.5 family, designed for production-grade visual creation and complex image-editing workflows. It supports text-to-image generation and image-to-image editing with strong prompt adherence, realistic imagery, consistent visual structure, and precise edits that preserve unaffected regions. Compared with the standard MAI-Image-2.5 model, the Pro variant adds stronger object consistency across complex scenes, character consistency across different views and moments, improved material and physical-property accuracy, and enhanced spatial and geometric reasoning. It is optimized for visually dense compositions, commercial imagery, concept art, product visualization, advertising, character-focused content, production design, and other professional workflows where image fidelity and consistency are prioritized.
|
oneinfer MAI-Image-2.6 is Microsoft's flagship image generation and editing model designed for high-quality, design-ready visual creation and precise professional image editing. It supports text-to-image generation, image-to-image editing with multiple visual references, web-grounded generation using current information, and model-directed dynamic aspect-ratio selection. The model improves text rendering, portraits, 3D imagery, photorealism, commercial imagery, product visualization, branding, and cinematic content while providing precise control over composition, style, structure, and detail. It is designed for professional creative workflows where maximum image quality, prompt fidelity, visual consistency, and controllable editing are prioritized.
|
oneinfer MAI-Image-2.6-Flash is Microsoft's fast and cost-efficient image generation and editing model in the MAI-Image-2.6 family, optimized for latency-sensitive and high-throughput production workloads. It delivers image quality comparable to the flagship MAI-Image-2.6 model while prioritizing generation speed and serving efficiency. The model supports text-to-image generation, image-to-image editing with multiple reference images, precise edits that preserve unaffected regions, web-grounded image generation using current information from Bing Search, and dynamic model-selected aspect ratios. It supports outputs up to approximately 1.5K resolution and is designed for advertising, product imagery, concept visualization, production design, creative content, image editing, and large-scale generative media applications.
|
 openai GPT Image 2 is OpenAI's image generation and editing model from April 2026, the generation before GPT Image 2.5. It takes flexible sizes up to 4K and is billed by the text and image tokens each request uses.
|
openai GPT-Image-2.5 Flare is OpenAI's fast, high-quality image generation and editing model optimized for everyday creative applications and high-volume workloads. It provides improved image quality, instruction following, subject preservation, editing consistency, natural lighting, and richer textures while delivering substantially lower generation latency. Flare is designed for creator content, social media, product experiences, visual search, rapid image prototyping, iterative editing, and applications where speed and image quality are both important.
|
openai GPT-Image-2.5 Sunburst is OpenAI's most capable image generation and editing model, designed for premium visual workflows that require high precision, detailed instruction following, and tight control across edits. It delivers sharper details, natural lighting, richer textures, improved subject preservation, stronger layout and text rendering, and reliable multi-turn image editing. Sunburst is optimized for production-ready creative work, polished product imagery, advertising assets, and other complex visual tasks where generation quality and precision are prioritized over latency.
|
openai GPT Image 2.6 Flare is the faster of OpenAI's two GPT Image 2.6 models, from November 2026, for generating and editing images. It takes sizes up to 4K and is billed by the text and image tokens each request uses.
|
openai GPT Image 2.6 Sunburst is OpenAI's GPT Image 2.6 model for detailed images and precise edits, from November 2026. It takes sizes up to 4K and is billed by the text and image tokens each request uses.
|
How to choose
- Marketing creative
- Product imagery
- Design iteration
- Image editing
Pricing methodology
OneInfer compares only positive prices with the same billing unit. Per-minute, per-character, per-token, per-image, per-video, and per-second rates remain separate. Prices can change, so the current model page and console remain the source of truth.
Frequently asked questions
How are image model prices compared?
OneInfer compares models using a stated per-image rate. Resolution or quality tiers are shown separately when they affect price.
Do all image models support editing?
No. Generation and editing are separate capabilities and are displayed only when they are present in verified model metadata.