Antigravity supports a wide range of AI models including FLUX, Stable Diffusion, Imagen, Ideogram, and Recraft for image generation, plus video, audio, and language models. This article breaks down every supported model, what each one excels at, and how to pick the right one for your workflow.
Antigravity has been building a reputation as one of the more versatile AI platforms available today, and a big part of that reputation comes from its model support. Not every AI tool gives you access to the top-tier models across every category. Antigravity does. Before you choose it as your primary AI creation tool, knowing exactly which models are under the hood will save you time and help you set the right expectations.
This article breaks down every major AI model Antigravity supports, from image generators to language models, with a focus on what each one is actually good at and where each one falls short.
Antigravity is not a model in itself. It is a platform that routes your creative requests to different best-in-class AI models depending on your task. Think of it as a dispatcher: you describe what you need, and Antigravity connects you to the right specialized model to produce it.
This architecture is increasingly common among serious AI tools because no single model dominates every category. FLUX might win on photorealism while Ideogram leads on text-in-image accuracy. Stable Diffusion offers fine-tuning flexibility that a closed model cannot match. By supporting multiple models, Antigravity avoids the weakness of betting everything on one provider.
The value of any AI platform is directly tied to the quality of models it can access. A platform built only on one provider is fragile: when that provider updates or degrades a model, every user suffers. Multi-model platforms like Antigravity can route around these issues by switching to a different model in the same category.
💡 For users, this means more consistent results over time, and the ability to choose the model that fits your specific style or requirement rather than accepting a one-size-fits-all output.
Image generation is the largest category Antigravity supports, and the breadth here is impressive. Here is how the major model families break down.
The FLUX family from Black Forest Labs has become the standard against which other image generators are measured. Antigravity supports multiple FLUX variants, each optimized for a different use case.
Flux.1 Dev is the developer-focused version, offering high-quality output with more controllability. It takes slightly longer to generate but the results are sharper and more prompt-accurate than most alternatives.
Flux Schnell trades some detail for dramatically faster inference. If you are iterating on a concept and need to preview many variations quickly, Schnell is the model you want. It produces clean, usable images in a fraction of the time.
Flux Pro sits at the top of the original lineup. The rendering quality on complex scenes, fabric textures, and facial details is noticeably better than Schnell, making it the right choice for final outputs.
Flux 1.1 Pro Ultra pushes the ceiling further, generating up to 4-megapixel images with stunning resolution. When you need print-ready assets or large-format renders, this is the version that delivers.
| Model | Best For | Speed |
|---|---|---|
| Flux Schnell | Fast iteration, previews | Very Fast |
| Flux.1 Dev | Controlled, detailed output | Moderate |
| Flux Pro | High-quality final renders | Moderate |
| Flux 1.1 Pro Ultra | 4MP resolution assets | Slower |
The newer Flux 2 Pro and Flux 2 Max versions extend this further, adding improved instruction-following and higher native resolution support. Antigravity's integration of the full FLUX 2 family puts it ahead of platforms that only support one or two variants.
Stability AI's Stable Diffusion remains the backbone of the open-source image generation world. Antigravity supports multiple versions across the SD family.
Stable Diffusion 3.5 Large is the flagship of the current lineup. Its triple-flow architecture handles complex prompts with multiple subjects significantly better than older SD versions, and the color accuracy is exceptional.
Stable Diffusion 3.5 Large Turbo uses distillation techniques to cut generation time in half while preserving most of the quality. For production workflows where time matters, this is a strong option.
Stable Diffusion 3.5 Medium is the community-friendly version with a smaller computational footprint. It is faster and cheaper to run while still producing results that outperform older SDXL generations.
The classic SDXL and SDXL Lightning 4Step remain available for workflows that rely on SDXL-trained LoRAs and ControlNets. The existing ecosystem of fine-tuned SDXL models is enormous, and Antigravity does not lock you out of it.
💡 Power tip: If you have a specific art style or subject domain you work in repeatedly (fashion, architecture, product shots), SDXL with a domain-specific LoRA will consistently outperform general-purpose models on your specific use case.
Google's Imagen family has established itself as a top-tier contender for photorealistic output. Antigravity has integrated all three current versions.
Imagen 3 introduced richer lighting simulation and better spatial reasoning than its predecessors. Complex scenes with multiple light sources, reflections, and shadows are rendered more accurately.
Imagen 4 significantly improved fine detail rendering, particularly in textures like fabric, hair, and natural materials. Side-by-side comparisons with competing models show Imagen 4 winning on these micro-detail categories.
Imagen 4 Ultra is the maximum-quality version, optimized for demanding professional applications. Imagen 4 Fast provides a speed-optimized path when you need Imagen-quality aesthetics without the full generation time.
One of the oldest pain points in AI image generation is getting text to render correctly inside an image. Ideogram was built specifically to solve this problem, and Antigravity's support for the full Ideogram lineup is a significant advantage for designers and marketers.
Ideogram v3 Quality is the current premium option, delivering accurate, sharp text rendering inside complex visual compositions. Logos, signage, product labels, and typography-heavy creative assets are where this model shines.
Ideogram v3 Turbo and Ideogram v3 Balanced give you the same core text accuracy at faster inference speeds, trading some compositional detail for throughput.
Older versions including Ideogram v2 and Ideogram v2 Turbo remain supported for compatibility with existing workflows.
💡 When to use Ideogram: Any time your prompt includes specific words, brand names, or numbers that must appear correctly in the output. Do not use Flux or SD for this, as text accuracy is not their strength.
Recraft v4 is unique because it can produce vector-compatible output through Recraft v4 SVG and Recraft v4 Pro SVG, giving designers editable, scalable assets rather than fixed-pixel images. The Recraft v4 Pro tier adds higher resolution and print-ready output specifications.
For any workflow that produces brand assets, icons, illustrations, or materials that will be used across different sizes and formats, Recraft's SVG output is a capability that most competing platforms cannot match.
Antigravity's video model support covers both text-to-video and image-to-video generation, two distinct workflows with different use cases.
Text-to-video in Antigravity allows you to describe a scene in plain language and receive a short video clip. The models available span different resolutions, durations, and motion styles. Quality varies significantly between models in this category, and the right choice depends heavily on your content type.
For cinematic motion with realistic physics, the WAN video family produces some of the best results currently available. For short social-media clips where speed matters more than length, lighter-weight models offer faster turnaround with acceptable quality.
Image-to-video is arguably more practically useful than text-to-video for many commercial applications. You start with a still image (from FLUX, Imagen, or any other source) and the model animates it into a short clip. This workflow is particularly effective for product marketing, where you might generate a perfect product shot with FLUX and then animate it with subtle camera movement for a social ad.
Antigravity is not limited to visual AI. Its LLM support includes access to large language models for text generation, content writing, summarization, and conversation. These models handle the text-based side of creative workflows: writing copy for generated images, generating prompt variations, summarizing long documents, and more.
The combination of LLM and image generation in a single platform removes workflow friction. You can write a brief, use the LLM to expand it into detailed image prompts, then feed those prompts directly to FLUX or Imagen without switching tools.
Vision-language models in Antigravity can analyze images and generate descriptive text. This is useful for:
The ability to flow from image analysis back into text prompts creates a feedback loop that skilled creators use to refine their outputs over time.
Beyond generation, Antigravity includes several utility models that solve common post-processing problems.
Super resolution models take an existing image and upscale it 2x to 4x while adding realistic detail rather than simply stretching pixels. This is critical for any workflow where images need to go from web resolution to print resolution, or from standard definition to 4K display size.
These models use different approaches: some analyze image content to generate contextually appropriate detail, while others use diffusion-based techniques to add texture that matches the original style. The choice between them depends on the subject matter and the type of detail you need.
Background removal in Antigravity handles complex edges including hair, fur, transparent objects, and fine fabric textures with significantly higher accuracy than traditional algorithmic cutout tools. This is particularly valuable for:
Text-to-speech models in Antigravity convert written content into natural-sounding voice audio. Use cases range from video narration to accessibility tools to content localization across languages.
Speech-to-text transcription enables workflows where you want to capture spoken ideas and convert them to structured text for further processing. AI music generation allows you to describe a mood or style and receive a backing track, useful for video projects where licensed music is a concern.
Every category in Antigravity presents a speed-quality spectrum. Understanding where you are in the creative process determines where on that spectrum you should operate.
| Stage | Priority | Recommended Approach |
|---|---|---|
| Early concept | Speed | Flux Schnell, SD 3.5 Medium, Ideogram v2 Turbo |
| Refinement | Balance | Flux Pro, Ideogram v3 Balanced |
| Final output | Quality | Flux 1.1 Pro Ultra, Imagen 4 Ultra, Ideogram v3 Quality |
Using the highest-quality model for every step is wasteful and slow. Using speed-optimized models for final outputs means delivering work that does not represent your best capability. The most efficient workflows match the model tier to the stage of work.
💡 Mix and match: Generate a draft with Schnell, refine the prompt, then run the final version with Flux 1.1 Pro Ultra. This cuts cost and time while maximizing final output quality.
The depth of model support in Antigravity is genuinely impressive, but reading about models is not the same as using them. The only way to know which model produces the results you need for your specific work is to test them directly.
Every model mentioned in this article, from Flux 1.1 Pro Ultra to Ideogram v3 Quality to Imagen 4 Ultra, is available on Picasso IA. You can access the full model catalog, run side-by-side comparisons, and build workflows that use different models for different stages of the same project.
Start with a prompt you know well, run it through three different models, and compare the outputs. The differences will immediately clarify which model fits your creative style and workflow requirements.
The model landscape changes fast. New versions of FLUX, Imagen, and Ideogram release regularly. Platforms that stay current on model updates give you access to improvements without requiring any workflow changes on your end. With Picasso IA, the model library is continuously updated. When a new version of your favorite model ships, it is available to you immediately, without reinstalling anything or changing your setup.
