Runway Absorbs MiniMax H3, Betting on a Multi-Model Creative Platform

MiniMax H3, a 2K open-weight video model with native audio, lands on Runway as the platform doubles down on its multi-model strategy

·
·
AuthorRunway
Read2 min
TopicVideo · Api
  • MiniMax H3 is live on Runway: The new open-weight video model is now accessible inside Runway's platform alongside all other frontier models.
  • H3 specs: Native 2K (2560x1440) resolution, up to 15-second clips at 24 FPS, synchronized stereo audio, and character consistency across scenes.
  • Multimodal inputs: A single H3 request accepts up to 9 reference images, 3 video clips, and 3 audio tracks for combined context generation.
  • Cost efficiency: A 15-second 2K clip costs roughly $1, and H3 ranks #1 on Artificial Analysis for video editing with audio.
  • Runway's platform play: Runway now bundles Seedance 2.0, Kling 3.0, Sora 2 Pro, and more alongside its own Gen-4.5, positioning itself as a multi-model creative hub.
  • Open weights: H3 weights are publicly available for self-hosting, private deployment, and fine-tuning via the MiniMax platform.

Runway just added MiniMax H3 to its growing roster of frontier video models, making the newly launched model available directly inside its creative platform the same day MiniMax officially released it. For creators, that means no new account, no new subscription, and no new workflow to learn. H3 is simply there, alongside everything else.

The addition is a small announcement with a larger story behind it. Runway is quietly repositioning itself from a video model company into something closer to a creative operating system, and MiniMax H3 is the latest piece of that puzzle.

What H3 Actually Does

MiniMax H3 generates clips up to 15 seconds long at 2560x1440 resolution with native audio and consistent characters across scenes. That combination, resolution plus audio in a single generation pass, is what sets it apart from most of the field.

H3 renders at full 2K pixel density inside the model itself, so crisp detail is the output rather than the result of a separate enhancement pass. In practice, that matters for anyone delivering to a client or publishing on a large display, where upscaled footage tends to fall apart.

  • Resolution: 5-15 second clips at 24 FPS with native stereo audio and a 1440p mode.
  • Inputs: One H3 API request accepts up to nine reference images, three video clips, and three audio tracks, capped at twelve files.
  • Editing: Video-to-video editing works from simple natural-language instructions in early demos.

Keep reading

Don't miss what's next in AI

Join 300,000+ engineers and researchers who get the signal, not the noise. Create a free account to read the rest of this story.

  • Full access to in-depth AI research breakdowns
  • Be the first to know what's trending before it hits mainstream
  • Daily curated papers, repos, and industry moves