Model family · MiniMax

MiniMax Video Models on OneInfer

The MiniMax video family spans the open-weight H3 foundation, fal's faster H3 Max post-train and the Hailuo generation line. Choose by resolution, editing needs and latency.

MiniMax video family

ModelRoleResolutionDurationPage
MiniMax H3 MaxFast, adherence-tuned H3 post-train480p / 768p5–15sView →
MiniMax H3Open-weight multimodal foundation with editingUp to 2K5–15sView →
MiniMax Hailuo 2.3Cinematic hosted generationUp to 1080p6–10sView →
MiniMax Hailuo 2.3 FastLower-latency Hailuo tierUp to 1080p6–10sView →
MiniMax Hailuo 02Earlier Hailuo generationUp to 1080pUp to 10sView →

How to choose

Use H3 Max for fast iteration and adherence at up to 768p. Use base H3 when you need open weights, 2K delivery or instruction-based editing. Use the Hailuo line when an existing production workflow already targets its endpoints.

Ready to test the workflow?

Create account & add credits

H3 Max is a distinct checkpoint

Do not treat H3 Max as an alias for MiniMax H3. fal post-trained it separately and publishes different resolution, performance and endpoint constraints.

Frequently asked questions

Which MiniMax video model should I use?

Start with H3 Max for fast 480p or 768p generation. Choose base H3 for 2K, open weights or video editing.

Put MiniMax H3 Max to work

Fund a controlled evaluation, send a reference frame or document, and measure quality and cost on your own workload.