Model comparison

🎨 GPT Image 2.5

Flare

Text-to-image, references, and mask editing. Manage jobs with Studio Unified or keep your OpenAI Images SDK.

sizeAuto / tiers / pixel sizes6quality options, including auto

🎨 GPT Image 2.5

Sunburst

Uses the same request structure as Flare. Swap the model name to compare results.

sizeAuto / tiers / pixel sizes6quality options, including auto

🎨 GPT Image

🎨 GPT Image 2

Generation and editing with size and quality controls.

sizeAuto / tiers / pixel sizes4quality options, including auto

🍌 Gemini 3.1 Flash Image

🍌 Nano Banana 2

Use Google's native multimodal structure or call it through Studio Unified.

1K · 2K · 4KOutput resolutionaspect ratioLandscape, portrait, or square

🍌 Gemini 3 Pro Image

🍌 Nano Banana Pro

Text, images, and iterative editing, with resolution and aspect-ratio controls.

1K · 2K · 4KOutput resolutionaspect ratioLandscape, portrait, or square

🍌 Gemini Nano Banana 2.1

🍌 Nano Banana 2.1

The new Nano Banana model, available through native multimodal requests and Studio Unified.

1K · 2K · 4KOutput resolutionaspect ratioLandscape, portrait, or square

gpt-image-2.5 is an alias for Flare. Use the full model ID in new configurations.

Which model name belongs in each protocol?

Studio Unifiedopenai/gpt-image-2openai/gpt-image-2.5-flareopenai/gpt-image-2.5-sunburstgoogle/gemini-3.1-flash-imagegoogle/gemini-3-pro-imagegoogle/gemini-nano-banana-2.1

OpenAI Imagesgpt-image-2gpt-image-2.5-flaregpt-image-2.5-sunburst

Gemini Nativegemini-3.1-flash-imagegemini-3-pro-imagegemini-nano-banana-2.1

A practical way to choose

01Choose the protocol

Use Studio Unified for a common surface; keep an existing OpenAI or Gemini integration on its native protocol.

02Match the task

Compare materials and text for products, consistency for people, and composition for posters. Run a small controlled comparison.

03Set a default

Record resolution, aspect ratio, references, and prompt together before setting a workflow default.

Studio Unified model names

Studio Unified uses provider prefixes:

openai/gpt-image-2
openai/gpt-image-2.5-flare
openai/gpt-image-2.5-sunburst
openai/gpt-image-2.6
openai/gpt-image-2.6-flare
openai/gpt-image-2.6-sunburst
google/gemini-3.1-flash-image
google/gemini-3-pro-image
google/gemini-nano-banana-2.1

Use the unprefixed API model names in OpenAI Images and Gemini Native.

Parameter differences

🎨 GPT Image uses size and quality. Image 2 supports auto, low, medium, and high; 2.5 Flare and Sunburst also support xhigh and max. Use a base model name and a size supported by the selected channel; no size suffix is needed. The bare 2.6 name resolves to Flare. Models, supported tiers, automatic sizing, and prices come from the current image catalog.

🍌 Nano Banana uses resolution and aspect_ratio in Studio Unified, and generationConfig.imageConfig or generationConfig.responseFormat.image in Gemini Native.

Choosing a model

For product images, compare materials, text, and branding. For characters, compare faces, outfits, and consistency. For slides and posters, compare typography and composition. Keep the prompt, references, and resolution comparable, then record the output and latency before choosing your workflow's default model.