Updated

Pika H3 Flex

Pika announced H3 Flex, a post-trained version of MiniMax H3 for iterative video creation. Pika says it supports up to 12 multimodal references, synchronized audio, and open weights for local use.

By No Code MBA · Facts checked

Provider
Pika
Announcement date
Status at last check
Generally available

What changed

Pika says it post-trained MiniMax H3 to generate in six steps instead of 50 and tuned it for creative control. H3 Flex accepts up to 12 image, video, and audio references; Pika says the API can provide five separate audio stems. The company also released generator adapters and inference code for local use.

Access and availability

Pika says H3 Flex is available through Pika API Club. It also released generator adapters and inference code for local use under the model's license terms. Pika says the model is planned for its creation product on November 2, 2026.

Official access documentation

What you can build

For a short video workflow, compare a small reference-guided generation with your current process, and account separately for local setup and Pika-hosted finishing features.

Limitations

Pika's seven-second figure measures transformer computation for a 15-second 768p clip, not total API latency. The 1080p to 4K upsampler, audio-stem separation, optional prompt enhancement, and accelerated serving are Pika-hosted features. Local performance depends on hardware and setup. Leaderboard placement and performance statements are Pika's launch-time claims.

Sources and release history