# Pika releases Pika H3 Flex, an open-weights video model with synchronized audio in six sampling steps

Canonical: https://theagenticweekly.com/news/pika-h3-flex-open-weights-video-model · Evidence: Reported · First seen: 2026-11-01 00:18 UTC · Source published: not stated by the source · Brief published: 2026-11-01 00:22 UTC · Updated: 2026-11-01 00:22 UTC · Author: Agentic Weekly

> Pika released Pika H3 Flex on October 29, 2026, an open-weights video model it says "ranks #1 among open-source text-to-video models". Post-trained from MiniMax H3, it generates video with synchronized stereo audio in six sampling steps and accepts up to 12 reference inputs.

Models in this brief: [Pika H3 Flex](https://theagenticweekly.com/models/pika-h3-flex)

## What changed

- Pika H3 Flex is a post-trained version of MiniMax H3; the released weights are adapters distributed under the MiniMax H3 Community License.
- It generates video with synchronized stereo audio in six sampling steps and accepts up to 12 reference inputs for shot direction.
- Pika says a 15-second 768p clip with audio can take as little as seven seconds of transformer compute, and notes that is not the full request time.
- Local inference runs at native 480p or 768p; Pika's hosted API adds 1080p, 2K and 4K, separate audio stems and prompt enhancement.

## In the source's words

> "an open-source omni video model family that ranks #1 among open-source text-to-video models"
> (Pika: H3 Flex technical deep dive, https://pika.art/blog/pika-h3-flex-research)

> "It builds on MiniMax-H3 and generates video with synchronized stereo audio in six sampling steps."
> (Pika H3 Flex model card (Hugging Face), https://huggingface.co/Pika-Labs/Pika-H3-Flex)

## What we have not verified

- The #1 ranking and the up-to-91% comparison are Pika's own evaluations; we did not run the model.
- Speed figures depend on hardware Pika did not fully specify in the quoted line.

## Timeline

- Source published: not stated by the source
- First seen by our watcher: 2026-11-01 00:18 UTC (via Hugging Face trending models (editorial release sweep))
- Brief published: 2026-11-01 00:22 UTC

## Sources

- [Pika blog: H3 Flex technical deep dive](https://pika.art/blog/pika-h3-flex-research) (primary source) (fetched 2026-10-31)
- [Hugging Face model card](https://huggingface.co/Pika-Labs/Pika-H3-Flex) (primary source) (fetched 2026-10-31)
- [Pika H3 Flex product page](https://pika.art/pika-h3-flex) (primary source) (fetched 2026-10-31)

## Related

- [Pika H3 Flex: model facts and sources](https://theagenticweekly.com/models/pika-h3-flex)
- [All models: compare price, context and evidence](https://theagenticweekly.com/models)
- [Pika: verdict and setup guide](https://theagenticweekly.com/tools/pika)
- [Alternatives to Pika](https://theagenticweekly.com/alternatives/pika)
- [All news briefs](https://theagenticweekly.com/news)
- [This week's releases](https://theagenticweekly.com/this-week)
- [New AI models this week](https://theagenticweekly.com/models/this-week)
- [AI model releases by month](https://theagenticweekly.com/models/releases)
- [Best LLM for coding](https://theagenticweekly.com/best/llm-for-coding)

## How to cite

Agentic Weekly, "Pika releases Pika H3 Flex, an open-weights video model with synchronized audio in six sampling steps", published 2026-11-01 00:22 UTC, https://theagenticweekly.com/news/pika-h3-flex-open-weights-video-model