Recraft V3 is the hosted image model known as Red Panda; Recraft V4 is a separate, later generation. Both support raster and vector output, but V3 retains API operations such as inpainting and exact text placement that V4 does not support. Recraft describes these as proprietary models, with no open-weight download in the cited access documentation. Model identities V3 operations V4 operations
Use Recraft V3 as the product name when researching access, features, or integration. The codename is useful for recognizing the model's history; the API identifier is what belongs in a request.
What are the key facts about Recraft V3, also called Red Panda?
| Field | Verified detail |
|---|---|
| Developer | Recraft. Overview |
| Released | October 30, 2024. Announcement |
| Type | Proprietary hosted image generation with raster and vector options. Overview |
| Size or parameters | Not published in the cited V3 model reference. V3 reference |
| License and access | Proprietary service; Recraft app and API access, with no open-weight distribution. Overview Announcement |
| Where it runs | Recraft's hosted platform and external API. Endpoint reference |
The three names to retain are Red Panda for the codename, Recraft V3 for the model generation, and recraftv3 for a raster API request. Overview V3 reference
This mapping helps when comparing an old saved result with a new run. The application name alone does not identify which model produced a particular image.
Which design controls does Recraft V3 provide?
Recraft's V3 announcement emphasizes text generation, composition, and style control. It also describes placing text in a design and generating vector artwork. Release announcement
Those capabilities make the model relevant to design briefs with explicit structure. A useful brief specifies the message, the main visual, and their relationship, rather than describing only an artistic mood.
For example, write a required headline separately from the illustration description. Then decide which element should attract attention first. Review the output against that hierarchy before considering decorative details.
A model can interpret a visual brief attractively while missing its purpose. If the audience must read the headline immediately, assess that requirement at the intended display size.
The announcement also explains style input as a way to guide visual treatment without retraining the model. That distinction matters when planning a branded series. Announcement
Start by choosing what should remain consistent across the series: palette, visual texture, composition, or another clear property. Avoid asking a single vague style label to carry every requirement.
What limits matter when choosing Recraft V3 or V4?
The Red Panda name does not imply a separate downloadable checkpoint. Recraft's official model taxonomy places it under V3 within a proprietary family. Overview
Likewise, a historical model reputation does not answer a current production question. Use the actual brief, output format, and finishing requirements to decide whether a model is suitable.
The release announcement describes a leaderboard based on pairwise image preferences. Such an evaluation concerns the images and comparison setup used; it should not be converted into a promised success rate for your project. Announcement
Recraft's API appendix limits V3 prompts to 1,000 characters. Keep essential content clear and compact rather than copying an entire creative strategy into the generation field. API appendix
A later model generation also need not support the same operations. Consult the operation list for the model you intend to call, especially when the workflow includes editing after the initial image. V3 V4
How do you select Recraft V3 explicitly in an API request?
Use the official API to request V3 explicitly. Recraft's setup page requires a positive API units balance before generating a token in your profile; then send a request with the documented model identifier. API setup Endpoint reference
curl --fail-with-body https://external.api.recraft.ai/v1/images/generations \
-H "Authorization: Bearer $RECRAFT_API_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"model": "recraftv3",
"prompt": "A bookshop poster reading OPEN LATE, warm interior illustration, clear headline",
"n": 1
}'
Set RECRAFT_API_TOKEN before running the example. The prompt is an original test brief, while the authentication and request shape follow the official reference. Endpoints
Download the returned image and retain the request. Label the file with the model generation so a future comparison does not depend on remembering which interface was open.
The sibling Recraft V3 API and SVG guide covers output selection and a more detailed integration workflow.
For this identity-focused evaluation, create a small comparison sheet. Include the model identifier, prompt, output type, and the particular requirement you want to assess.
Review text separately from illustration. Mark a misspelled word as a text failure even when the overall design is attractive; mark an unsuitable composition separately so revisions have a clear target.
Keep the original output when making corrections. This lets you estimate how much work the generator saved and which part of the design still needed human decisions.
How does Recraft V3 compare with Recraft V4?
Recraft V4 was released in February 2026. Its reference lists standard, Pro, and vector models, with a different set of supported operations and controls. V4 documentation
| Question | Recraft V3 | Recraft V4 |
|---|---|---|
| Is this the Red Panda generation? | Yes. Overview | No; it is the later V4 generation. V4 |
| Is vector generation documented? | Yes. V3 | Yes. V4 |
| What deserves compatibility review? | Text-layout and editing operations. V3 | Its listed generation and image-to-image operations. V4 |
Recraft's documentation identifies V3-specific text-positioning and editing support. Therefore, migration should include checking each operation, not just changing the model name in an otherwise unchanged integration. V3
The sibling Recraft V4 prompting guide provides a separate angle on that generation.
For a useful comparison, use the same creative requirement but respect each model's documented inputs. Report the settings used rather than claiming a controlled test while silently enabling different helper features.
If your deciding requirement is local execution, consult the ComfyUI pillar. Establish weight availability and runtime support before comparing hardware needs.
What else should you know about Red Panda and Recraft versions?
Who made the Red Panda image model?
Recraft developed the Red Panda image model and identifies it as Recraft V3 in its API documentation. Use the official V3 model reference when checking supported features or API access. Overview
Which API model name corresponds to Red Panda?
The Red Panda generation is Recraft V3: use recraftv3 for raster images or recraftv3_vector for SVG output. The codename itself is not a supported value in the endpoint's model list. Model identities Identifiers
Is Recraft V3 the same as Recraft V4?
Recraft V3 and Recraft V4 are different model generations with different operation support. Check each model's operation list and identifier when reproducing results or updating an integration. V3 V4
Does the Red Panda codename establish image quality?
Red Panda identifies Recraft V3 and does not establish whether a particular output meets your design brief. Compare the result with your requirements for wording, composition, format, and finishing effort.
Sources
- Recraft model taxonomy and Red Panda identity
- Recraft V3 release announcement
- Recraft V3 model reference
- Recraft API endpoints
- Recraft API appendix
- Recraft V4 model reference
Top comments (0)