Model providers

OpenAI's image model sets the bar for instruction following — when a composition must obey a detailed spec exactly, GPT Image 2 is the reference.

GPT Image 2 · GPT Image 2.5 Flare · GPT Image 2.5 Sunburst · GPT Image 2 HD →

Alibaba's Qwen Image and Z-Image models cover text-to-image and reference-image edits, offered here as variants with looser content filtering.

Qwen Image 2.1 (Unrestricted) · Qwen Image 2.0 · Qwen Image 2.0 Pro · Qwen Image 3.0 · Qwen Image 3.0 Pro · Wan 2.7 Image · Z-Image Turbo · HappyHorse 1.1 · HappyHorse 1.0 · Wan 3.0 Video · Wan 3.0 Video Prime →

ByteDance's cloud arm publishes the Seedream image family and the Seedance video family — Seedream 5 Pro (Unrestricted) for detailed stills and Seedance 2.0 for short clips with sound.

Seedream 5 Pro (Unrestricted) · Seedream 5 Pro · Seedance 2.0 · Seedance 2.0 4K · Seedance 2.0 Fast · Seedance 2.0 Mini · Seedance 2.5 →

NovelAI's image models specialise in anime and illustration styles.

NovelAI V5 (Unrestricted) · NovelAI V4.5 (Unrestricted) →

Google's Nano Banana image models handle both generation and conversational editing, from fast drafts to high-fidelity renders.

Nano Banana 2.1 · Nano Banana 3 Pro · Nano Banana 3.1 Flash →

xAI's Grok Image models generate images from text through the same images API as every other model here.

Grok Image 2.0 →

Kling 3.0 Image · Kling 3.0 · Kling 3.0 Omni · Kling O1 · Kling Omni 4.0 →

MiniMax H3 (768p) →