Wan 3.0, Nano Banana 2.1, MiniMax H3 and 4 more new models on Varosity

As of October 25, 2026, seven new models are on Varosity: Wan 3.0, MiniMax H3, Grok Imagine Video 1.5 and Seedance 2.0 Mini for video, and Nano Banana 2.1, Seedream 5.0 Pro and Qwen Image 3.0 for images. Each one is billed at its provider's list price with no markup, on Varosity credits (1 credit = 1¢; credits are charged in whole cents, so an image priced at $0.045 costs 5¢). We rendered every one of them live on Varosity on launch day; the prices and timings below are from those runs.

Video

You pay per second of the clip, in whole seconds; a request below a model's minimum is rounded up to it (2 seconds on MiniMax H3 renders and bills 4). A 2-second Wan 3.0 clip cost 20¢ in our tests and took about 70 seconds to render; a 4-second MiniMax H3 clip cost 40¢ and took 1.5–3 minutes; a 4-second Grok Imagine Video 1.5 clip cost 56¢; a 4-second Seedance 2.0 Mini clip cost 48¢.

• Wan 3.0 is the one for long single takes: up to 30 seconds in one clip, with generated audio, in 16:9, 9:16 or 1:1.
• MiniMax H3 is the one of these four that does 21:9 widescreen from text. Its documentation lists no audio setting, but every clip we rendered came with a soundtrack.
• Grok Imagine Video 1.5 is text-to-video only on Varosity, in 16:9, 9:16 or 1:1. For Grok image-to-video, use the earlier grok-imagine-video.
• Seedance 2.0 Mini animates a start frame at half the price of Seedance 2.0 Fast. It needs an image; without one the request is refused before anything is charged.

Images

• Nano Banana 2.1 is the pick for designs with text in them. In our test it set "OPEN LATE" cleanly on a poster, then kept the lighthouse and the lettering when we asked for the same scene on a snowy night.
• Seedream 5.0 Pro follows detailed scene prompts closely, and its edits keep the room and subject while changing what you ask for.
• Qwen Image 3.0 gives you 2K images (2730×1535 at 16:9) for 3¢. Its signs and lettering come out as made-up characters, so use Nano Banana 2.1 when the text has to be readable.

Image calls are synchronous and take longer than you might expect: 26–45 seconds for a new image and up to about 80 seconds for an edit in our tests. Keep the request open rather than retrying.

Try them

From an agent connected over MCP (Claude, ChatGPT, Gemini, Cursor, Claude Code and others — setup), just ask:

Use Varosity to make a 5-second Wan 3.0 clip of a fisherman hauling a glowing net onto a boat at dawn.

Or call the API directly:

curl -X POST https://varosity.ai/api/v1/video/generate \
  -H "Authorization: Bearer $VAROSITY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"modelId":"wan-3.0","prompt":"A fisherman hauls a glowing net onto a wooden boat at dawn, mist on the water","aspectRatio":"16:9","durationSec":5}'

curl -X POST https://varosity.ai/api/v1/images \
  -H "Authorization: Bearer $VAROSITY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"nano-banana-2.1","prompt":"A minimalist poster of a lighthouse at dusk with the words \"Open Late\"","aspect_ratio":"16:9"}'

Every model's page lists its exact price, lengths, shapes and prompting tips, and the full catalog is at varosity.ai/models. Most new accounts include a small starter credit, and top-ups start at $5.