The right modelfor every task.

Claude tuned for narrative and creative writing.

Anthropic's flagship — the hardest reasoning and longest tasks.

Balanced Claude: close to Opus quality at everyday speed.

Large-context general model.

xAI's newest Grok model.

MiniMax M3 with a 1M context window.

The cheap tier of the family, for cost-sensitive work.

Frontier model for complex professional work.

Balances intelligence and cost.

Standard tier: the provider's own content filter stays on.

Premium tier: the provider's platform filter is disabled.

Anime SDXL on our own GPUs.

OpenAI's image model, dependable at text inside pictures and at edits.

Sharper detail and closer prompt adherence than gpt-image-1.

The quick, low-cost tier — for drafts and iterating.

OpenAI's latest image model, strongest at typography.

A pinned snapshot of gpt-image-2, for output that stays reproducible.

Tier de precision de GPT Image 2.5: edicion fiel a costa de mas tiempo de generacion.

Natural-language model on our own GPUs.

Photorealistic SDXL with an upscale-and-refine pass and optional detailers.

MiniMax image generation, with character reference support.

Alibaba's image model, strong on English and Chinese text.

The higher-fidelity Qwen 2.0 tier.

Qwen's newer generation, with better composition and detail.

The top Qwen tier, for final-quality renders.

Qwen image editing without the content filter.

Qwen Image without the content filter.

Photorealistic SDXL on our own GPUs.

ByteDance's high-resolution image model, with reference-image support.

ByteDance's latest Seedream generation.

Illustrious SDXL on our own GPUs.

Fast Seedance for quick drafts. 4–12s, up to 1080p.

Dreamina's newer Seedance, with steadier motion. 4–12s, up to 1080p.

MiniMax H3 video, 4-15s at 768P or 2K.

ByteDance video with image-to-video. 4–12s, token-metered.

First and last frame video, billed by the second.

All-in-one video: text, image, video and audio reference, up to 30s with native sound.

All-in-one video: text, image, video and audio reference, up to 30s with native sound. No content filter.

Wan's top video tier. Up to 30s at 1080p, billed by the second.

Wan's top video tier. Up to 30s at 1080p, billed by the second. No content filter.

Wan's fastest tier at 1080p, 2K or 4K, up to 30s with native sound. No content filter.

All-in-one video at 1080p, 2K or 4K, up to 30s with native sound. No content filter.

xAI's expressive voice, with 20 languages detected from the text.

MiniMax's highest-fidelity voice. $100 per million characters.

Faster and cheaper than HD. $60 per million characters.