Understand this story
On November 1, 2026, Pika released Pika H3 Flex, a derivative model based on MiniMax H3, and opened the adapter weights and inference code. According to Leiphone, the model jointly generates visuals and sound through six-step denoising, includes two sets of LoRA adapters, natively supports 480P and 768P output, can generate videos of about 5 to 15 seconds with synchronized stereo audio, and offers a ComfyUI workflow and a Hugging Face Diffusers integration method. The report states that the team first used supervised fine-tuning, then compressed 50-step denoising to 6 steps using Projected Distribution Matching Distillation (PDMD), with high-resolution enhancement and audio track separation provided by the Pika API. There are currently no facts of subsequent developments.
AI-generated from public reports
Latest development · Report
From Nvidia to Pika, Why Silicon Valley Companies Build on MiniMax H3Developments
1 confirmed developments
- Source:English text is being prepared.
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 3
36Kr · AI
From Nvidia to Pika, Why Silicon Valley Companies Build on MiniMax H3
After MiniMax opened the weights of its multimodal model H3 in August this year, Nvidia's research team released Open-H3-Regenerate-2K based on H3, Pika launched the post-trained version H3 Flex on October 29, and fal trained H3 Max. The article states that within one month of H3's open-sourcing, downloads exceeded 24 million, spawning more than 300 public models, and in September it ranked first on the Arena image-to-video leaderboard with a score of 1495. The author argues that open weights turn partners into nodes through which model vendors enter more industries, and cites the H3 community license agreement, noting that commercial products or services with annual revenue exceeding $20 million must obtain prior written authorization from MiniMax.
Nov 1
Leiphone
Pika Open-Sources H3 Flex Based on MiniMax H3, Achieving Joint Audio-Video Generation in Six Steps
Pika has released Pika H3 Flex, a derivative model of MiniMax H3, opening the adapter weights and inference code, and jointly generating visuals and sound through six-step denoising. The model includes two sets of LoRA adapters, natively supports 480P and 768P output, can generate approximately 5 to 15 seconds of video with synchronized stereo audio, and provides a ComfyUI workflow and a Hugging Face Diffusers integration method. The team first used supervised fine-tuning, then used Projected Distribution Matching Distillation (PDMD) to compress 50-step denoising to 6 steps; high-resolution enhancement and audio track separation are provided by the Pika API.