Models
Every model you can use, what it's good at, and what it costs.
Last verified
Overview
Batch Banana has image and video models from several labs. Each one draws differently and costs a different number of credits. This page lists them all, in the model menu's default order.
The tables below come from the same list the app reads, so they always match the model menu. Select a model in the menu to see the same facts on its card, plus a longer description, its speed and whether it can work from your own images.
Not sure where to start? Which model for what gives a first pick for each kind of job, then runs five families on the same eight prompts.
Image models
| Model | Best for | Credits per image | Reference images |
|---|---|---|---|
| GPT Image 2.5 Sunburst | OpenAI's most capable image model, built for precise edits | 30 | Up to 8 |
| GPT Image 2.5 Flare | OpenAI's fastest model for everyday image generation | 30 | Up to 8 |
| Nano Banana 2.1 | Pro quality at Flash speed, up to 4K | 10 | Up to 14 |
| Nano Banana 2 Lite | Google's fastest and cheapest, built for iterating | 8 | Up to 14 |
| Nano Banana Pro | Highest quality from Google, up to 4K | 32 | Up to 14 |
| Flux 3 Image | Generation and editing in one model, with web search | 6 | Up to 10 |
| Seedream 5 Pro | ByteDance's flagship with sharp 1K and 2K output | 11 | Up to 10 |
| Seedream 5 Lite | Built-in reasoning with example-based editing up to 3K | 10 | Up to 14 |
| Krea 2 | Photorealism and expressive artistic styles | 14 | Up to 10 |
| Recraft V4.1 Pro | 2x resolution Recraft for print and large-scale work | 59 | None |
| Recraft V4.1 Pro SVG | High-resolution SVG vectors for large-scale production | 59 | None |
| Arrow 2 | Clean, editable SVG vectors from text | 59 | Up to 14 |
| Ideogram 4.5 | Best-in-class text and graphic design, with precise edits | 14 | Up to 5 |
| Recraft V4.1 | Art-directed images with precise text rendering | 10 | None |
| Recraft V4.1 SVG | Native editable SVG vectors from text prompts | 10 | None |
| Flux 2 Klein 9B | Enhanced realism, crisper text and better color adherence | 4 | Up to 5 |
| Flux 2 Klein 4B | Precise modifications using natural language descriptions | 1 | Up to 5 |
| Imagen 4 Ultra | Maximum quality Imagen with highest fidelity | 14 | None |
| Flux Kontext Max | Complex scenes with strong spatial awareness | 19 | Up to 8 |
| Imagen 4 | Photorealistic results with accurate text rendering | 10 | None |
Credits per image is the price at the model's default setting. Models with a Resolution or Quality setting cost a different amount at other steps; Credit costs lists every step.
Reference images is how many of your own images the model reads alongside the prompt. A model with none works from words alone.
Video models
| Model | Credits per second | Length | Sound | Reference images |
|---|---|---|---|---|
| Seedance 2.5 | 54 | 4 to 30 s | Optional | Up to 30 |
| MiniMax H3 | 19 | 4 to 15 s | Always | Up to 9 |
| Seedance 2.0 | 42 | 4 to 15 s | Optional | Up to 9 |
| Flux 3 Video | 40 | 5 to 20 s | Optional | Up to 8 |
| Veo 3.1 | 94 | 4, 6, 8 s | Optional | Up to 3 |
| Gemini Omni 1.1 Flash | 35 | 10 s | Always | Up to 3 |
| Kling 3.0 | 53 | 3 to 15 s | Optional | None |
| Runway Gen-4.5 | 28 | 5, 10 s | No | None |
A video's price depends on its length, its resolution and whether it has sound. The figure here is the base rate per second; Credit costs gives the price of a full clip.
Prompting guides
Seven model families have their own guide, with their makers' advice, our own tests and example images:
- Nano Banana: Google's models, from quick drafts to 4K.
- GPT Image: exact text, careful edits and layouts.
- Ideogram: posters, logos and anything with words on it.
- Flux: lettering in many languages, web search for real places, edits from up to 10 images.
- Seedream: drafts, edits copied by example and precise layouts.
- Recraft: designed images and real SVG files from words alone.
- Arrow: real SVG vectors for icons and marks.