Pika launches H3 Flex, an open-weight video model built on MiniMax H3
Pika’s seven-second benchmark measures transformer computation for a 15-second 768p clip, not total API request time.
TLDR
On October 29, Pika introduced H3 Flex, a post-trained MiniMax H3 video model with public adapters and inference code for 480p and 768p generation. Pika also offers hosted 4K finishing and separate audio stems. Its advertised seven-second result measures transformer computation for a 15-second 768p clip in a controlled test, not the time a user waits for a finished API request.
Combined views
15.6K
20 Sources, first seen ago

