The Open Weights
LatestModelsLeaderboardsCompanies
Subscribe
The Open Weights

The daily record of open-source AI. New model releases, leaderboards, and what's coming next — written for people who ship.

Refreshed every 12 hours

Discover

  • Latest releases
  • New today
  • Trending models

Browse

  • All models
  • Companies
  • Categories
  • Leaderboards

About

  • About
  • Editorial policy
  • Hardware estimates
  • RSS feed
  • llms.txt
  • Newsletter

© 2026 The Open Weights. An independent publication.

PrivacyTermsSMSAggregated by Claude · curated by humans.
LatestIFM2
IFMImage → Video

IFM releases PAN2, an open world-simulation video model

The Institute of Foundation Models ships a permissively licensed diffusion-transformer system for turning text and images into video.

Oct 7, 2026
NotableApache 2.0
PAN2

The Institute of Foundation Models has published PAN2, a diffusion-transformer model for generating video from text prompts and still images. The release is positioned as a world-simulation system, meaning it aims to model how scenes evolve over time rather than simply stitch frames together.

PAN2 supports two primary workflows: text-to-video, where a written prompt produces a clip, and image-to-video, where a single still is animated into motion. The latter is listed as the model's primary modality, suggesting IFM sees conditioning on a reference frame as the central use case.

Why it matters

The most consequential detail may be the license. PAN2 ships under Apache-2.0, one of the most permissive terms available, which allows commercial use, modification, and redistribution with minimal restrictions. That stands apart from many video models that arrive under research-only or custom non-commercial terms.

  • Modalities: text-to-video and image-to-video
  • Architecture: diffusion transformer
  • License: Apache-2.0
  • Framing: world simulation

IFM has not published parameter counts, context details, or benchmark figures alongside the release, so the practical quality and hardware requirements remain to be tested by the community. For now, the model and its details are available on the Hugging Face repository.

Sources

  • IFM/PAN2

    Hugging Face

    Visit
OlderNVIDIA's Nemotron 3.5 Transcribe Brings Promptable ASRNVIDIA · Speech → Text · last monthNewerQwen-Image-2.1-Turbo targets faster generationQwen · Alibaba · Text → Image · last month

Get the model

Hugging Face

Specs

Size67.4 GB
PrecisionBF16
LicenseAPACHE-2.0
Downloads593
Likes342

Can you run it?

Runs on a 32 GB GPU at FP8.

  • BF16 (as published)

    80 GB GPU (A100 / H100) · 128 GB Mac Studio

    68.4 GB
  • FP8

    32 GB GPU (RTX 5090) · 48 GB Mac

    30.2 GB
Your machine
  • Covers the published checkpoint; text encoders and VAE loaded alongside may add a few GB (many runtimes offload them to CPU).

Based on Hugging Face weights metadata. Assumes an 8K context and ~1 GB runtime overhead; actual needs vary. How we estimate


Modalities

Image → VideoText → Video

The Weekly Weights

Every open release that mattered, one email a week.

0 comments

No comments yet. Be the first to weigh in.

More from IFM

All IFM releases →
K2-Horizon-375B-A23B
IFM/Text / LLMMulti-GPU server

IFM releases K2-Horizon, a 375B open-weight MoE

The flagship model uses a mixture-of-experts design that activates just 23 billion parameters per token, keeping inference costs in check.

Sep 1, 2026
K2-Horizon-375B-A23B
IFM/Text / LLMRuns on a laptop

IFM releases K2-Horizon-7B with open pretraining data

The dense 7-billion-parameter model ships as open weights alongside the datasets used to train it.

Sep 1, 2026
K2-Horizon-MoVA-36B-A4B
IFM/Text / LLMConsumer GPU

IFM's K2-Horizon MoVA Ships as a 36B MoE Model

The open-weight language model activates just 4B of its 36B parameters per token, aiming for efficiency without shedding capacity.

Sep 1, 2026

More in Image → Video

All Image → Video →
LongCat-Video-Avatar-1.6
Meituan/Image → Video

Meituan's LongCat Adds Audio-Driven Avatar Video

The LongCat team releases an image-to-video model that animates still portraits from audio input.

Nov 13, 2026
NVIDIA/Image → Video

NVIDIA Releases Cosmos 3 for World Simulation

The new open models push NVIDIA's world-foundation effort toward interactive, real-time video generation.

Nov 5, 2026
NVIDIA/Text → Video

NVIDIA's Cosmos 3 Brings Open Interactive World Models

The new open release pushes beyond passive video generation toward simulated worlds that respond to input.

Nov 5, 2026