MBZUAI’s IFM Releases PAN 2, an Open 29B World Model for Physical AI

The Institute of Foundation Models (IFM) at MBZUAI released PAN 2 , an open world model for physical intelligence. The PAN 2 world model takes a scene and a text action and simulates the outcome as video. IFM announced it on X on November 10, 2026. Weights, training code and inference code ship under Apache 2.0 . For robotics and agent developers, the pitch is simple: test a plan in simulation before acting on it. TL;DR Size: 29B-parameter diffusion transformer (MMDiT), dense. Inputs: text plus an optional first frame. Output: 704×1248 video at 24 fps, up to 20 s. Runs on: Data-center GPUs. Each worker holds the full 29B model in BF16, with block offloading as fallback. Performance: Leads 6 tested models on full plans. Trails MiniMax-H3 on single actions and human preference. Best: 34.0% full-plan completion on 100 multi-step tasks, the highest of all models tested (self-reported). Worst: About 2 in 3 plans miss a step. Physics-IQ system default: 37.48 vs 42.91 for MiniMax-H3. Bottom line: Best: a fully open world model that leads on multi-step plans. Worst: long plans still fail often, and hardware needs are high. What is PAN 2? PAN 2 is a general-purpose world model from IFM that simulates how a scene changes after an action. It follows the Generative Latent Prediction (GLP) architecture from IFM’s Critique of World Models paper. It succeeds PAN , released in November 2025. The goal is simulative reasoning . An agent proposes plans, PAN 2 simulates each one, and the agent acts on the best. How does PAN 2 work? PAN 2 splits simulation into 2 kinds of reasoning : Conceptual reasoning : An LLM reads the observation and the requested actions. It writes the next world state as structured JSON: events, timed actions bound to actors, and camera motion. Any instruction-following LLM can fill this role. IFM’s report uses Kimi K3 by default. Perceptual reasoning : A 29B MMDiT predicts the video latents. IFM trained it from scratch on hundreds of millions of annotated videos. A frozen Qwen3.5-9B text encoder conditions it. The Wan 2.2 VAE decoder renders the final frames. IFM’s ablation shows the LLM matters. With no conceptual reasoning, full-plan completion drops from 34.0% to 0.0%. Step completion falls from 75.1% to 8.7%. A format-only rewrite recovers part of the gap, reaching 27.8% full-plan completion. IFM also dropped gameplay, animation and simulation footage after early training. A 9B test model trained without it won 58% of blinded votes. Prev</button> <button class="btn pri" id="mtpNext">Next </button> <button class="btn" id="mtpPlay"> Auto-play</button> <span class="hint">Click any step to jump</span> </div> </div> <div class="pane" id="p-bench"> <div class="seg" id="mtpSeg"></div> <div class="bars" id="mtpBars"></div> <p class="note" id="mtpNote"></p> </div> <div class="pane" id="p-abl"> <div class="seg" id="mtpAblSeg"> <button class="chip" data-a="0">None: action sent straight to diffusion</button> <button class="chip" data-a="1">Format-only rewrite</button> <button class="chip on" data-a="2">Full conceptual reasoning</button> </div> <div class="abl"> <div class="card"><div class="k">6 capabilities</div><div class="v" id="ab0">84.6<small>%</small></div><div class="mini"><i id="ib0"></i></div></div> <div class="card"><div class="k">Step completion</div><div class="v" id="ab1">75.1<small>%</small></div><div class="mini"><i id="ib1"></i></div></div> <div class="card"><div class="k">Full-plan completion</div><div class="v" id="ab2">34.0<small>%</small></div><div class="mini"><i id="ib2"></i></div></div> </div> <p class="note">IFM ablation with Kimi K3 as the reasoner; the diffusion model stays fixed. Without the LLM, no multi-step plan completes. Source: IFM technical blog.</p> </div> <div class="pane" id="p-run"> <div class="grid"> <div class="spec"><div class="k">Generator</div><div class="v">29B MMDiT, BF16, trained from scratch</div></div> <div class="spec"><div class="k">Tasks</div><div class="v">Text-to-video, image-to-video</div></div> <div class="spec"><div class="k">Output</div><div class="v">704x1248 at 24 fps, up to 20 s (481 frames)</div></div> <div class="spec"><div class="k">Default clip</div><div class="v">241 frames, about 10 s</div></div> <div class="spec"><div class="k">Sampler</div><div class="v">50 steps · flow shift 9 · CFG 2.5</div></div> <div class="spec"><div class="k">Caption budget</div><div class="v">1,536 tokens hard limit</div></div> <div class="spec"><div class="k">Serving</div><div class="v">Diffusers · vLLM · FastVideo · generate.py</div></div> <div class="spec"><div class="k">GPU memory</div><div class="v">Full 29B per worker; block offload fallback; VRAM not disclosed</div></div> <div class="spec"><div class="k">Safety</div><div class="v">pan2-guardrail: unsafe T2V 33.9% → 1.5%</div></div> <div class="spec"><div class="k">License</div><div class="v">Apache 2.0: weights, training + inference code</div></div> </div> <pre>hf download IFM/PAN2 --include "transformer/*" hf download Qwen/Qwen3.5-9B hf download Wan-AI/Wan2.2-TI2V-5B-Diffusers</pre> </div> <div class="ft"> <span>Sources: <a href="https://huggingface.co/IFM/PAN2" target="_blank" rel="noopener">HF model card</a> · <a href="https://ifm.ai/blog/pan2/" target="_blank" rel="noopener">IFM blog</a> · scores self-reported</span> <span class="mtp">© Marktechpost</span> </div> </div> <script> (function(){ var R=document.getElementById('mtp-pan2'); function post(){try{parent.postMessage({type:'mtp-pan2-h',h:R.offsetHeight+40},'*')}catch(e){}} // tabs R.querySelectorAll('.tab').forEach(function(b){b.addEventListener('click',function(){ R.querySelectorAll('.tab').forEach(function(x){x.classList.remove('on')}); R.querySelectorAll('.pane').forEach(function(x){x.classList.remove('on')}); b.classList.add('on');document.getElementById('p-'+b.dataset.p).classList.add('on'); if(b.dataset.p==='bench')drawBars(cur); if(b.dataset.p==='abl')setAbl(abl); setTimeout(post,60); })}); // pipeline var D=[ ['Observe: current world + action','Input is an optional starting image plus text, and actions described in text with a duration. Example: "Accelerate mangrove forest growth to protect the coast."',''], ['Conceptual reasoning backbone (LLM)','An instruction-followi