inclusionAI releases Duo 1.0 causal-base, a full-duplex multimodal world model initialized from MiniMax H3
inclusionAI has published the Duo 1.0 causal-base weights on Hugging Face. Duo is described as a full-duplex multimodal world model for real-time audio-visual interaction, initialized from MiniMax H3 and trained further by the Duo team.
inclusionAI has released the weights for Duo 1.0 causal-base on Hugging Face. The model card describes Duo as a full-duplex multimodal world model for real-time audio-visual interaction, and the repository subtitle calls it "A Reciprocal World Model for Full-Duplex Interaction."
What is in the release
The repository holds the Duo 1.0 causal-base variant as a complete model bundle. Per the model card:
- Version: Duo 1.0
- Variant: Causal base
- Package: Complete model bundle
- Weight format: PyTorch Distributed Checkpoint (DCP)
- Default sampler: Euler, 50 steps, flow shift 1.0
The bundle contains the Realization weights and backbone, the Intent model and adapter, video and audio codecs, a speech recognizer, and a forced aligner. Runtime integrations can read realization/complete.json to identify the packaged model family and weight format.
Base model and license
According to the model card, the weights were initialized from MiniMax H3 and then trained by the Duo team. Applicable terms are in the LICENSE and NOTICE files in the repository.
Access
The weights can be downloaded with hf download inclusionAI/Duo-1.0-causal-base --local-dir weights. Code, configuration and runtime-specific integration instructions are in the separate inclusionAI/Duo repository. As of November 13, 2026, no Inference Provider deploys the model on Hugging Face.
The model belongs to a Hugging Face collection that inclusionAI describes as "a modular Reciprocal World Model for full-duplex interaction." The collection has 4 items and was updated 13 minutes before the page was read. A linked paper, arXiv 2611.15131, was published two days earlier.
What this means
Duo ships as a modular system rather than a single checkpoint: the world model, intent components, codecs, speech recognizer and forced aligner are packaged together. Teams that want to run it will start from the bundle and the runtime instructions in the inclusionAI/Duo repository. With no hosted provider listed yet, local deployment from the DCP weights is the only access route shown on the model page.
Related Articles
Huawei Cloud to unveil CloudRobo-WM1.2, its first Ascend-native world model, on November 16
Huawei Cloud will launch CloudRobo-WM1.2, an embodied world model for robots, at an online event on November 16 (14:00 to 15:30 Beijing time). Its poster bills the model as the first Ascend-native world model from the Industry AI Dream Factory. Parameter counts, benchmarks, availability and pricing have not been disclosed.
Perplexity releases Decider v1.3 and Decider Flash v1.3 with text-span extraction and Apache-2.0 weights
Perplexity released Decider v1.3 and Decider Flash v1.3 on November 12, 2026, adding verbatim text-span extraction and downloadable Apache-2.0 weights. The company reports 91.6% mean accuracy for the larger model across 63 query-classification tasks.
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.
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...