Pika Drops H3 Flex, an Open Video Model Generating Audio 1.8x Faster

Takeaways

• Pika released Pika H3 Flex , an open-weight video+audio model post-trained on MiniMax H3.
• Generates 15-second clips with soundtrack in as little as 7 seconds using a custom 6-step sampling schedule.
• Ranks #7 on DesignArena t2v with Elo 1293, the top open-weight entry and above H3Max and Seedance 2.5.
• Supports up to 12 reference inputs (9 images, 3 videos, 3 audio) plus 5 exportable audio stems via API.
• Weights on Hugging Face , inference code on GitHub , with ComfyUI and diffusers support day one.
• API pricing starts at $0.01/sec at launch; local inference needs an H100 80GB and ~215GB of downloads.