AI News: New ChatGPT & Google Models (14 Updates Worth Knowing)
Open-weight video model up to 4K, faster than real time, with soundtracks; 220 GB weights needing an H100-class GPU. Tested free on a Hugging Face space, about 30 seconds per clip.
Pika's open-weight video model post-trained on MiniMax H3, generating synchronized video and audio up to 4K.
Updated EU status checked Last shown
Pika H3 Flex at a glance
- What is Pika H3 Flex?
- Pika H3 Flex is a model from Pika Labs. Pika's open-weight video model post-trained on MiniMax H3, generating synchronized video and audio up to 4K.
- Who makes Pika H3 Flex?
- Pika H3 Flex is made by Pika Labs.
- Does Pika H3 Flex work in the EU?
- Available in the EU. Pika's terms contain no EU exclusion and reference the EU Digital Services Act; the weights are open and downloadable. No explicit supported-countries list was found. Last verified 14 Nov 2026.
- How much does Pika H3 Flex cost?
- Freemium. Hosted: Free plan with credit packs; Starter $10/mo (600 credits), Creator $35/mo (2,600 credits), Fancy $95/mo (8,550 credits); Creator and Fancy include free daily H3 Flex generations (limited time). Weights are reportedly downloadable for local use.
- Where was Pika H3 Flex shown?
- In 2 creator videos by Matt Wolfe, Matthew Berman. Most recently by Matt Wolfe on 6 Nov 2026, at 9:30 in “AI News: New ChatGPT & Google Models (14 Updates Worth Knowing)”.
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
About
Pika H3 Flex is a video generation model from Pika Labs, post-trained from MiniMax H3 and described by Pika as built for creative iteration. It is an omni-generator that produces synchronized video and stereo audio of up to 15 seconds, using six denoising steps, about 8× fewer than its 50-step teacher model. Pika's announcement says it accepts up to 12 references, outputs five separate audio stems, and can finish at up to 4K.\n\nPika states that the weights are open. YouTube creators report about 220 GB of weights needing an H100-class GPU, a roughly 7-second generation time for a 15-second clip, and a free test on a Hugging Face space at about 30 seconds per clip. These figures come from creators and are not confirmed on the provided Pika pages.\n\nIt is also available in Pika's hosted platform (Video Studio), where the pricing page lists it at 768p, 5s. Creator and Fancy plan subscribers get free H3 Flex generations every day, a limited-time offer. Limits: the license terms and the exact weight repository were not provided in the source material.
Key features
- Synchronized video and stereo audio generation up to 15 seconds
- Up to 12 reference inputs
- Five separate audio stems
- Finishing at up to 4K
- Six-step distilled denoising, about 8× fewer steps than the 50-step teacher
- Open weights; also available in the Pika web platform
2 moments
As seen on YouTube
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Kimi K3.1 weights OUT, open source computer use, SOTA image model, science model, Pika H3: AI NEWS
Open-weight text-to-video model trained on MiniMax H3, ranked number one open-weight on Design Arena. It supports up to 4K, audio stems and multi-asset inputs, makes a 15-second clip in about 7 seconds, and costs about 1 cent per second. The creator says he uses it locally.
Keep exploring
Similar AI tools
Closest neighbours by meaning, not just by category.
FO
FastH3 Omni
Model · Hao AI Lab
Distilled MiniMax H3 reference-to-video-and-audio model from Hao AI Lab, generating in 8 steps instead of 50.
Available in EUVideo Generation
20 Oct 2026Open source
MH
MiniMax H3
Model · MiniMax
Open-weight video generation model from MiniMax, part of its H3 family alongside H3 Image.
UnverifiedVideo Generation
Release date n/a
FA
FastH3
Model · Hao AI Lab
Open-weight 4-step sparse-distilled version of MiniMax H3 for fast text-to-video-and-audio generation on Blackwell GPUs.
Available in EUVideo Generation
Release date n/aOpen source
PI
P-H3 Image
Model · Pruna AI
Open-weight image generation and editing model from Pruna AI, distilled from MiniMax's H3 video model to run in 8 steps.
Available in EUImage Generation
Release date n/aOpen source
MH
MiniMax H3 Image
Model · MiniMax
Open-weights MiniMax image generation and editing model producing up to ~16 MP images in 6 sampling steps.
Available in EUImage Generation
3 Nov 2026Open source
MH
MiniMax H3 audio generation
Research · MiniMax
Open-weight multimodal model from MiniMax whose video pipeline can be run audio-only to generate music and sound effects.
UnverifiedVoice & Audio
Release date n/a