Reference · image generation · external source

Image models on OpenRouter, compared

Ten image-generation models: nine routed by the OpenRouter API, plus xAI's Grok Imagine Image 2.0 (available via grok.com/imagine and the xAI API). They are compared on the facts published about each: price, context window and provider. Four are featured, three of them with a sample output. The OpenRouter nine were checked against OpenRouter on 2026-10-22; the Grok entry cites xAI's announcement. Everything carries a numbered marker to the page it came from. This is a referenced comparison, not a ranking — the arena itself ranks video models and still has no votes.

Compare all ten models
Featured · samples served locally where published

Featured models

Four featured models: the three on OpenRouter's featured compare view, each shown with a generation from its own model page, downloaded and served locally, plus xAI's Grok Imagine Image 2.0, featured without a sample image but with links to run it. The metric rows are the published figures for that model, cited per row.

Sample generation from the Nano Banana 2 model page on OpenRouter
Sample generation from the OpenRouter model page.

Nano Banana 2 (Gemini 3.1 Flash Image) Google

Google's state of the art image generation and editing model, delivering Pro-level visual quality at Flash speed. Combines advanced contextual understanding with fast, cost-efficient inference; aspect ratios can be controlled with the image_config API parameter. [3]

Released2026-06-18 [3]
Context131K tokens [3]
Price$0.50/M input tokens, $60/M output tokens [3]
ProvidersGoogle AI Studio, Google Vertex [3]

View on OpenRouter →

Sample generation from the Muse Image model page on OpenRouter
Sample generation from the OpenRouter model page.

Muse Image Meta

An agentic image generation model from Meta that generates and edits images from text and reference images. Unlike single-pass image models, it reasons before it renders, breaking down multi-part prompts and refining its output within the chain of thought, and invokes web search for factual accuracy on knowledge-intensive prompts. Supports text-to-image, targeted editing, multi-image composition, reference-image conditioning and iterative editing. [4]

Released2026-08-26 [4]
Context66K tokens [4]
Price$0.01/image [4]
ProvidersMeta [4]

View on OpenRouter →

Sample generation from the GPT Image 2.5 Sunburst model page on OpenRouter
Sample generation from the OpenRouter model page.

GPT Image 2.5 Sunburst OpenAI

An image generation and editing model from OpenAI, positioned as the precision-oriented tier of the GPT Image 2.5 series. Suited to detailed creative work where editing accuracy matters more than generation speed, via the dedicated Images API. [5]

Released2026-09-09 [5]
Context400K tokens [5]
Price$8/M input tokens, $30/M output tokens [5]
ProvidersOpenAI [5]

View on OpenRouter →

Grok Imagine Image 2.0 xAI

Precise image generation and editing from xAI, generally available as the new Quality Mode on grok.com/imagine and the iOS and Android apps. Follows instructions closely down to typography and layout; adds magic-wand region editing, segmentation, background removal and multi-ref editing with up to 5 input images, plus smart resize across ratios from 1:2 to 2:1. xAI states it ranks second in the world in both text-to-image generation and image editing (Arena boards, as of the 2026-08-07 announcement). [6]

Released2026-08-07 [6]
ContextNot stated [6]
PriceNot stated here (not routed by OpenRouter) [6]
ProvidersxAI (grok.com, iOS, Android) [6]

Open grok.com/imagine → · Try on the xAI API →

xAI announcement →

External benchmarks · OpenRouter’s own evaluation

OpenRouter benchmarks

OpenRouter runs a fixed prompt suite against the image models it routes and publishes how many checks each model passes, plus where that places it on overall quality, cost and speed. The 3 models with published figures are shown here; the other models on this page carry no published benchmark.

Nano Banana 2 (Gemini 3.1 Flash Image) Google

66 of 81 checks across 15 prompts [3]

Overall 14th of 55Cost 40th of 55 (cheapest first)Speed 9th of 55 (fastest first)

Muse Image Meta

71 of 81 checks across 15 prompts [4]

Overall 7th of 55Cost 4th of 55 (cheapest first)Speed 29th of 55 (fastest first)

GPT Image 2.5 Sunburst OpenAI

70 of 81 checks across 15 prompts [5]

Overall 5th of 55Cost 19th of 55 (cheapest first)Speed 33rd of 55 (fastest first)

OpenRouter’s own evaluation figures, quoted with attribution; checked 2026-10-22. Higher checks and lower rank numbers are better.

All ten models compared

Prefer hands-on evidence over a spec sheet? generate images across models on AIContentDrop — switching models there is a dropdown, not a new integration. Check the credit pricing for image models before you commit, or drive the same prompts from your own pipeline through the AIContentDrop REST API.

All 10 image-generation models — nine routed by OpenRouter, plus xAI's Grok Imagine Image 2.0 direct from xAI — OpenRouter figures checked 2026-10-22 [1]
ModelVendorReleased ContextPriceProviders
Nano Banana 2 (Gemini 3.1 Flash Image)[3]
View on OpenRouter →
Google2026-06-18131K tokens$0.50 / $60 per 1MGoogle AI Studio, Google Vertex
Muse Image[4]
View on OpenRouter →
Meta2026-08-2666K tokens$0.01 per imageMeta
GPT Image 2.5 Sunburst[5]
View on OpenRouter →
OpenAI2026-09-09400K tokens$8 / $30 per 1MOpenAI
Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)[7]
View on OpenRouter →
Google2026-06-3065K tokens$0.25 / $1.50 per 1MGoogle AI Studio, Google Vertex
Nano Banana Pro (Gemini 3 Pro Image)[8]
View on OpenRouter →
Google2026-06-18131K tokens$2 / $12 per 1MGoogle AI Studio, Google Vertex
GPT-5 Image[9]
View on OpenRouter →
OpenAI2025-10-14400K tokens$10 / $10 per 1MOpenAI
GPT-5 Image Mini[10]
View on OpenRouter →
OpenAI2025-10-16400K tokens$2.50 / $2 per 1MOpenAI
GPT-5.4 Image 2[11]
View on OpenRouter →
OpenAI2026-04-21272K tokens$8 / $15 per 1MOpenAI
Nano Banana (Gemini 2.5 Flash Image)[12]
View on OpenRouter →
Google2025-10-0732K tokens$0.30 / $2.50 per 1MGoogle AI Studio, Google Vertex
Grok Imagine Image 2.0[6]
xAI announcement →
xAI2026-08-07Not statednot on OpenRouterxAI (grok.com, iOS, Android)
External crowd boards · order only, no scores

Arena rankings

xAI’s Imagine Image 2.0 announcement quoted two crowd-voted Arena boards to position the model. The order is republished here as simple rank lists; the bars are decorative — longest is first place — and no underlying scores are shown.

Image Edit Arena

1gpt-image-2OpenAI
2grok-imagine-image-2 (low)xAI
3muse-imageMeta
4mai-image-2.5Microsoft AI
5grok-imagine-image-qualityxAI
6gemini-3-pro-image-2kGoogle
7chatgpt-image-high-fidelityOpenAI
8seedream-5.0-proByteDance

Text-to-Image Arena

1gpt-image-2OpenAI
2grok-imagine-image-2 (low)xAI
3reve-2.1Reve
4muse-imageMeta
5qwen-image-3.0-proAlibaba
6gemini-3.1-flash-imageGoogle
7seedream-5.0-proByteDance
8grok-imagine-image-qualityxAI

Source: Arena leaderboards as of 2026-08-07, via xAI’s Imagine Image 2.0 announcement[6]. Ordered ranking lists as published in xAI's announcement; Elo values are not republished, only the order. xAI models are listed on Arena under SpaceXAI.

Source and image credits

Source: OpenRouter (openrouter.ai)[2], checked 2026-10-22 · compare view: https://openrouter.ai/compare/google/gemini-3.1-flash-image/meta/muse-image/openai/gpt-image-2.5-sunburst[1] · xAI: https://x.ai/news/grok-imagine-image-2[6]. Image credits: Sample generations from the OpenRouter model pages.

FAQ

What is OpenRouter?

An API router that fronts many model providers behind one API and one account. For a developer, switching between the nine OpenRouter-routed models on this page is a slug change in the request, not a new integration — the same call shape reaches every provider they list. Grok Imagine Image 2.0 is not routed by OpenRouter; it is reached via grok.com/imagine and the xAI API instead.

Do these models make video?

No. Every model on this page is an image-generation model: text or reference images in, still images out. The arena's video model catalogue is a separate reference, and nothing on this page changes it — video voting is still pre-launch and no votes exist.

Are these your prices and sample images?

Neither. Prices, context windows and provider lists are published by OpenRouter and quoted here with attribution, and the sample images are generations from the OpenRouter model pages, downloaded and served locally rather than hotlinked. Every figure carries a numbered marker to the page it was read from.

Sources

  1. OpenRouter — image model compare view https://openrouter.ai/compare/google/gemini-3.1-flash-image/meta/muse-image/openai/gpt-image-2.5-sunburst — checked 2026-10-22
  2. OpenRouter — API router https://openrouter.ai — checked 2026-10-22
  3. OpenRouter — Nano Banana 2 (Gemini 3.1 Flash Image) (google/gemini-3.1-flash-image) https://openrouter.ai/google/gemini-3.1-flash-image — checked 2026-10-22
  4. OpenRouter — Muse Image (meta/muse-image) https://openrouter.ai/meta/muse-image — checked 2026-10-22
  5. OpenRouter — GPT Image 2.5 Sunburst (openai/gpt-image-2.5-sunburst) https://openrouter.ai/openai/gpt-image-2.5-sunburst — checked 2026-10-22
  6. xAI — Grok Imagine Image 2.0 announcement https://x.ai/news/grok-imagine-image-2 — checked 2026-10-22
  7. OpenRouter — Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) (google/gemini-3.1-flash-lite-image) https://openrouter.ai/google/gemini-3.1-flash-lite-image — checked 2026-10-22
  8. OpenRouter — Nano Banana Pro (Gemini 3 Pro Image) (google/gemini-3-pro-image) https://openrouter.ai/google/gemini-3-pro-image — checked 2026-10-22
  9. OpenRouter — GPT-5 Image (openai/gpt-5-image) https://openrouter.ai/openai/gpt-5-image — checked 2026-10-22
  10. OpenRouter — GPT-5 Image Mini (openai/gpt-5-image-mini) https://openrouter.ai/openai/gpt-5-image-mini — checked 2026-10-22
  11. OpenRouter — GPT-5.4 Image 2 (openai/gpt-5.4-image-2) https://openrouter.ai/openai/gpt-5.4-image-2 — checked 2026-10-22
  12. OpenRouter — Nano Banana (Gemini 2.5 Flash Image) (google/gemini-2.5-flash-image) https://openrouter.ai/google/gemini-2.5-flash-image — checked 2026-10-22

Each [n] above is the vendor page the fact beside the marker was read from, on the date shown. Where a vendor states nothing, the table says “Not stated” and no value is inferred from anywhere else.