Models that return images
These return pictures, and pictures are billed as image output tokens — one image costs several times what a paragraph of text costs. Check your balance before a bulk run: the hold placed on your wallet scales with the number of images you ask for.
11 models in this category · prices in toman per 1M tokens
| Model | Capabilities | Context | Input / 1M | Output / 1M | ≈ 1,000 words |
|---|---|---|---|---|---|
 Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)google/gemini-3.1-flash-lite-image | visionreasoningjsonimage out | 65,536 | 72,531 | 435,188 | 660 |
| Nano Banana (Gemini 2.5 Flash Image)google/gemini-2.5-flash-image | visionreasoningjsonimage out | 32,768 | 87,038 | 725,313 | 1,056 |
| Nano Banana 2 (Gemini 3.1 Flash Image)google/gemini-3.1-flash-image | visionreasoningjsonimage out | 131,072 | 145,063 | 870,375 | 1,320 |
| Nano Banana 2 (Gemini 3.1 Flash Image Preview)google/gemini-3.1-flash-image-preview | visionreasoningjsonimage out | 65,536 | 145,063 | 870,375 | 1,320 |
| GPT-5 Image Miniopenai/gpt-5-image-mini | visionreasoningjsonfiles | 400,000 | 725,313 | 580,250 | 1,697 |
| Nano Banana Pro (Gemini 3 Pro Image)google/gemini-3-pro-image | visiontoolsreasoningjson | 131,072 | 580,250 | 3,481,500 | 5,280 |
| Nano Banana Pro (Gemini 3 Pro Image Preview)google/gemini-3-pro-image-preview | visionreasoningjsonimage out | 65,536 | 580,250 | 3,481,500 | 5,280 |
| GPT-5 Imageopenai/gpt-5-image | visionreasoningjsonfiles | 400,000 | 2,901,250 | 2,901,250 | 7,543 |
| GPT-5.4 Image 2openai/gpt-5.4-image-2 | visionreasoningjsonfiles | 272,000 | 2,321,000 | 4,351,875 | 8,675 |
 Auto Routeropenrouter/auto | visiontoolsreasoningjson | 2,000,000 | — | — | — |
| Auto Router (Beta)openrouter/auto-beta | visiontoolsreasoningjson | 2,000,000 | — | — | — |
Other categories
Models that hold up in a real codebaseCheap models for high-volume workFree models, shared capacityModels that read imagesReasoning models, and what they costLong-context models for documents and codebasesModels that call your tools reliablyEmbedding models for semantic searchBatch variants (cheaper, slower)