> For the complete documentation index, see [llms.txt](https://docs.qolaba.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.qolaba.ai/image-generation/choosing-your-model.md). # Choosing Your Model Qolaba provides access to 15+ image generation models in one interface. Each model has distinct strengths across realism, artistic style, prompt comprehension, speed, and cost. Selecting the right model before generating directly affects output quality, available features, and credit usage. > **Page description:** A complete guide to all available image generation models in Qolaba — capabilities, credit costs, reference image support, and recommendations by use case. *** #### How Model Selection Works Each model has a **credit cost per image** displayed next to its name in the model selector. This cost, multiplied by your number of generations, gives you the total credit cost for that run. Beyond cost, models differ in: * **Output style** — photorealistic, artistic, illustrative, design-forward * **Prompt comprehension** — how accurately the model interprets detailed instructions * **Reference image support** — whether the model accepts uploaded images and how many * **Generation speed** — how quickly the model produces output *** #### Available Models | Model | Credits / Image | Reference Images | Best For | | ------------------- | --------------- | ---------------- | ----------------------------------------------------------- | | **Nano Banana Pro** | 49 | Up to 13 | Realistic visuals, polished creatives, high-quality renders | | **Nano Banana 2** | 25 | Up to 13 | Balanced quality and cost, general-purpose generation | | **GPT Image 2** | 18 | Up to 15 | Concept art, structured scenes, strong prompt comprehension | | **Flux 1.1 Pro** | 16 | Up to 8 | High-quality photorealistic and detailed outputs | | **Flux Dev** | 9 | Up to 8 | Rapid testing, cost-efficient drafts | | **Seedream 4.5** | 21 | — | Stylized visuals, artistic and expressive images | | **Recraft V3** | 22 | — | Vector-style, design-forward, brand-consistent visuals | | **Ideogram V3** | 22 | — | Text-in-image, typography, readable graphics | | **DALL-E 3** | 12 | — | Natural language comprehension, concept visualization | | **SD 3.5** | 23 | — | High-detail diffusion, strong composition | | **SD 3.5 Turbo** | 8 | — | Fast generation, good for drafts | | **SD 3.5 Medium** | 9 | — | Balanced speed and quality | | **SDXL** | 18 | — | Broad style range, established base model | | **ImageGen 4** | 4 | — | Quick tests, low-cost iterations | | **ImageGen Fast** | 3 | — | Most cost-effective, rapid iterations | *** #### Model Recommendations by Use Case **Highest quality photorealistic output** → **Nano Banana Pro** — Qolaba's premium model for polished, production-ready images. Best for client-facing creatives, high-fidelity product visuals, and any output where maximum realism matters. **General-purpose generation** → **Nano Banana 2** — The recommended starting point for most tasks. Balances quality, cost, and versatility across a wide range of prompt types and styles. **Strong prompt comprehension** → **GPT Image 2** — Handles complex, detailed instructions reliably. Best for structured scenes, multi-element compositions, and prompts that require precise interpretation. **Reference-guided generation** → **Nano Banana Pro or Nano Banana 2** — Both support up to 13 reference images. Use when visual consistency with existing assets is critical — brand imagery, character consistency, product styling. **Photorealistic detailed output** → **Flux 1.1 Pro** — Strong at producing highly detailed, photorealistic images. Good alternative to Nano Banana Pro at a lower credit cost. **Fast iteration and draft testing** → **Flux Dev, SD 3.5 Turbo, or ImageGen Fast** — Use these for prompt testing, concept exploration, and any workflow where speed and cost efficiency matter more than maximum quality. **Artistic and stylized output** → **Seedream 4.5** — Produces expressive, stylized visuals with an artistic quality. Well-suited for editorial imagery, creative campaigns, and illustrative content. **Design-forward and brand-consistent visuals** → **Recraft V3** — Excels at clean, vector-style outputs with strong design sensibility. Best for brand assets, UI illustrations, and design-forward content. **Text within images** → **Ideogram V3** — The strongest model for generating images that include readable, well-rendered text. Use for social media graphics, typographic compositions, and any image requiring legible text elements. **Concept visualization** → **DALL-E 3** — Strong natural language comprehension makes it effective for concept-driven prompts where the idea matters more than photorealism. **Most cost-effective** → **ImageGen Fast** at 3 credits per image — use for rapid iteration, high-volume testing, or any workflow where volume matters more than premium quality. *** #### Choosing Between Similar Models Some models are close in capability and cost — here is how to choose between them: **Nano Banana Pro vs. Nano Banana 2** Both support up to 13 reference images. Nano Banana Pro produces noticeably higher quality and more expressive outputs — use it for final production. Nano Banana 2 is better suited for drafts and general-purpose work where premium quality is not required. **Flux 1.1 Pro vs. Flux Dev** Flux 1.1 Pro produces higher quality photorealistic output. Flux Dev is significantly cheaper and faster — use it for testing prompts and validating composition before switching to Flux 1.1 Pro for final output. **SD 3.5 vs. SD 3.5 Turbo vs. SD 3.5 Medium** SD 3.5 produces the highest quality within the Stable Diffusion family. SD 3.5 Turbo is the fastest and cheapest — good for drafts. SD 3.5 Medium sits between the two in both quality and cost. **ImageGen 4 vs. ImageGen Fast** Both are low-cost options. ImageGen Fast is marginally cheaper and faster. Use either for rapid iteration — neither is intended for final production output. *** #### Reference Image Support by Model When your workflow requires uploading reference images for image-to-image generation, only specific models support this capability: | Model | Max Reference Images | | ------------------- | -------------------- | | **Nano Banana Pro** | 13 | | **Nano Banana 2** | 13 | | **GPT Image 2** | 15 | | **Flux 1.1 Pro** | 8 | | **Flux Dev** | 8 | All other models support text-to-image generation only. See Reference Image Uploads → for full guidance on uploading and using reference images. *** #### What's Next * Presets — System & Custom → * Input Controls → * Understanding Credit Usage → --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.qolaba.ai/image-generation/choosing-your-model.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.