Skip to main content
Recraft

Recraft v3

Recraft v3 is an image generation model from Recraft, a design-focused image generator known for text rendering inside images and style-controlled output.

From
$0.040 / image
across 1 provider

API Pricing

ProviderPrice /img
$0.040/img

Prices updated daily. Last check: Sep 25, 2026

Model Details

General

Creator
Recraft
Modalities
Text

Capabilities

Tool Calling
No
Open Source
No
Aliases
recraft-ai/recraft-v3

Strengths & Limitations

Strengths

  • Text-to-image generation aimed specifically at design output rather than generic image synthesis
  • Recraft positions the v3 generation around rendering legible text inside generated images, a common weak point for image models
  • Style control features intended to keep a consistent look across a set of generations
  • A vector-oriented variant of Recraft v3 is published for design work that needs scalable output
  • Available both through Recraft's own API and through third-party inference marketplaces under the `recraft-ai/recraft-v3` alias
  • Well suited to repeatable asset production such as icon sets, posters, and social creative

Limitations

  • Model weights are not distributed for self-hosting, so it cannot be run on your own GPUs
  • Priced per generated image rather than per token, which makes cost forecasting for high-volume pipelines depend on retry and iteration rates
  • Feature availability — including vector output and style controls — varies between Recraft's own API and third-party resellers
  • We do not track independent benchmark scores for this model, so quality claims should be validated on your own prompts
  • Design-oriented focus may be less relevant for photorealistic or scientific imagery workloads

Key Features

•Text-to-image generation
•Text rendering inside generated images
•Style controls for consistent output across a set of generations
•Vector-oriented output variant published alongside the raster model
•Aspect ratio and size options for design formats such as posters and social assets
•API access through Recraft plus third-party inference marketplaces
•Alias `recraft-ai/recraft-v3` used by hosted providers

About Recraft v3

Recraft v3 is a text-to-image generation model built by Recraft, a company whose products target graphic design and brand asset workflows rather than general-purpose image synthesis alone. It is offered through Recraft's own API and platform as well as through third-party inference marketplaces, where it is commonly referenced by the alias `recraft-ai/recraft-v3`. As an image_generation model, Recraft v3 is evaluated on prompt adherence, output resolution and aspect ratio control, stylistic consistency, and generation latency rather than on context window or token throughput. Recraft has positioned the v3 generation around rendering legible text within generated images and around style control, including the ability to apply a consistent look across multiple generations — features aimed at posters, ads, social assets, and illustration sets. Recraft also publishes a vector-oriented variant of v3 for design work that needs scalable output; capability details and variant availability differ by provider, so check the specific endpoint you plan to call. In practice, Recraft v3 is used by designers and product teams generating marketing creative, icon and illustration sets, and mockups where typography and a repeatable visual style matter. Compared with general-purpose diffusion models, its differentiator is the design-tooling angle — text-in-image and style governance — while general-purpose alternatives may offer broader ecosystem tooling, fine-tuning options, or open weights. Because image models are priced per image rather than per token, per-provider pricing structures vary; see the pricing table on this page.

Common Use Cases

Recraft v3 fits design and marketing asset production: posters, banners, ad creative, social graphics, icon and illustration sets, and product mockups. Its text-in-image emphasis makes it a candidate for creative that must include headlines, labels, or short copy directly in the render, which typically requires manual compositing with image models that garble type. The style-consistency features suit teams producing batches of assets that need to look like they belong together, such as an illustration library for a docs site or a campaign run across multiple formats. The vector-oriented variant is the relevant choice when output has to scale cleanly to logos, icons, or print. For photorealistic photography-style work, scientific visualization, or workloads that require self-hosted weights and fine-tuning, other image generators are likely a better match.

Frequently Asked Questions

How much does Recraft v3 cost?

Image models are billed per generated image, and the rate depends on which provider you use, the requested size or quality setting, and whether you are calling Recraft's own API or a third-party host. Rates change often, so check the pricing table on this page for current per-image pricing across providers.

What is Recraft v3 best used for?

Design-oriented image generation — marketing and social creative, posters, icon and illustration sets, and mockups — particularly where the image needs legible text rendered inside it or where a set of images must share a consistent style.

Can Recraft v3 produce vector graphics?

Recraft publishes a vector-oriented variant alongside the raster Recraft v3 model for work that needs scalable output such as icons and logos. Availability of that variant depends on the provider and endpoint you call, so confirm before building against it.

Can I run Recraft v3 on my own GPUs?

No. Recraft v3 is served through APIs — Recraft's own and third-party inference marketplaces, where it appears as `recraft-ai/recraft-v3` — and weights are not distributed for self-hosting. If self-hosting is a requirement, an open-weight image model is the appropriate category to compare.

How does Recraft v3 compare to general-purpose image models?

Its differentiator is the design-tooling angle: text rendering inside images, style control across generations, and a vector variant. General-purpose diffusion models often have broader community tooling, more fine-tuning and LoRA options, and in some cases open weights. Since we do not track independent benchmarks for Recraft v3, run your own prompt set against both before committing.