MBZUAI's IFM releases PAN 2, an Apache 2.0 world model with a 29B diffusion transformer
The Institute of Foundation Models (IFM) at MBZUAI released PAN 2, an Apache 2.0 world model. It couples an LLM planner with a 29B diffusion transformer to simulate up to 20 seconds of action outcomes as video.
The Institute of Foundation Models (IFM) at MBZUAI released PAN 2, an open world model under the Apache 2.0 license. According to MarkTechPost, PAN 2 couples an LLM planner with a 29B-parameter diffusion transformer and simulates up to 20 seconds of action outcomes as video.
What was released
- Model: PAN 2, described as a world model for physical AI
- Developer: Institute of Foundation Models (IFM) at MBZUAI
- License: Apache 2.0
- Architecture: an LLM planner paired with a 29B diffusion transformer
- Output: video simulating the outcomes of actions, up to 20 seconds long
Context window, API pricing, benchmark scores and training cutoff date are not yet disclosed.
How it works
PAN 2 splits the task into two components. The LLM planner handles the action side, and the 29B diffusion transformer renders the simulated outcome as video. The simulation horizon is up to 20 seconds per run.
What this means
PAN 2 pairs language-model planning with video-based simulation of action outcomes, and it ships under Apache 2.0. Teams working on physical AI can examine the two-part design and the 29B diffusion transformer directly. Performance comparisons will depend on benchmark results and technical documentation from IFM, which are not yet available here.
The report comes from MarkTechPost, published November 10, 2026.
Related Articles
Zyphra releases ZONOS3-Mini, an Apache 2.0 TTS model with 100M active parameters
Zyphra released ZONOS3-Mini, an Apache 2.0 text-to-speech model with 800M total parameters and 100M active parameters. It clones voices, steers delivery word by word and runs in real time in the browser.
MiniMax releases H3-Image: 33.1B-parameter model generates 16-megapixel images in 6 sampling steps
MiniMax has released MiniMax-H3-Image, a 33.1B-parameter diffusion transformer that handles text-to-image and image editing in one checkpoint. It generates images of about 16 megapixels directly, without an upscaler, using 6 sampling steps and no classifier-free guidance.
Ant Group's Robbyant open-sources LingBot-Touch, a tactile model trained on nearly 5,000 hours of data
Robbyant, the embodied-AI company under Ant Group, has open-sourced LingBot-Touch under Apache-2.0. The tactile foundation model was pretrained on nearly 5,000 hours of real data and covers magnetic, visuotactile and piezoresistive sensors. Tiny and Small checkpoints are on Hugging Face and ModelScope.
Fish Audio releases Live 3, a full-duplex voice model that reacts in as little as 186 ms
Fish Audio released Live 3, a full-duplex voice model for voice agents. It listens while it speaks, separates "mm-hmm" from real interruptions, and can react in as little as 186 ms. It is English only at launch.
Comments
Loading...