MiniMax · H3
MiniMax H3
Open weights, 2K video, native stereo.
What it is
MiniMax H3 is a general-purpose, omni-modal generative system: it reads text, images, video and audio as one context and generates video with native stereo sound. Clips run 4 to 15 seconds at 24fps, with a 2K mode that regenerates a 768p result in context rather than upscaling it.
Released in July 2026 and open-sourced in August, H3 is the model you can run yourself. Base checkpoints are published under a community licence, with serving recipes for SGLang, vLLM, Diffusers and ComfyUI; the hosted preprocessing and 2K regeneration stages remain API-only.
Profiles describe the model, not a guarantee: the Skill folder is what makes the work repeatable, and the agent's settings decide what the model can reach.
What it's best at
The work this engine is built for, and where it earns its place in a run.
Open weights
Base checkpoints under a community licence, deployable on your own GPUs.
Omni-modal context
Up to nine images, three videos and three audio clips — twelve files in total.
2K by regeneration
Re-renders its own 768p output with the original context to recover detail.
Native stereo
32 kHz stereo audio, with dialogue in eleven languages.
A good first run
Three video Skills to try on MiniMax H3 — each one a folder of instructions the model can pick up straight away.
Other models
The rest of the engine room.
ByteDance · Seedance 2.5
Seedance 2.5
ByteDance's latest video model: a 30-second audio-video scene from one pass, with up to 50 references.
- Duration
- 30s, extendable
- References
- 50 assets
ByteDance · Seedance 2.0
Seedance 2.0
ByteDance's unified audio-video model: text, image, audio and video references for 15-second multi-shot sequences.
- Duration
- 15s, multi-shot
- References
- 15 assets
Kuaishou · Kling 4.0
Kling 4.0
Kuaishou's production-grade model: 30-second takes, 10-bit HDR, omni-reference and precise keyframes.
- Duration
- 30s
- Output
- 4K HDR
Ready to extend your agent?
Connect once, then install any Skill from our studio straight from your agent.