Tencent Robotics X · robot AI model
Hy-Embodied-RxBrain-1.0
China · 2026
Sources retrieved: 9 November 2026 (A, B, C, D)
- Model type
- Embodied reasoningSource and note: Model typeB arxiv.org/html/2607.14187 · Retrieved 9 November 2026Note. The paper describes “a more unified architecture that enables embodied reasoning, world state prediction, and joint language-visual subgoal planning”. It sets the model apart from both vision-language models and generative world models: “Unlike vision-language models that emphasize scene understanding and textual decision making, or generative world models that mainly predict future visual states”.
- Inputs
- Camera LanguageSource and note: InputsA huggingface.co/tencent/Hy-Embodied-RxBrain-1.0 · Retrieved 9 November 2026Note. The model card lists “question answering and chain-of-thought over images and multi-frame video”.
- Access
- Open weights (download)Source and note: AccessA huggingface.co/tencent/Hy-Embodied-RxBrain-1.0 · Retrieved 9 November 2026Note. The model card says Tencent released “the technical report, official inference code, and model weights.”
- Parameters
- approximately 6.2BSource and note: ParametersA huggingface.co/tencent/Hy-Embodied-RxBrain-1.0 · Retrieved 9 November 2026Note. The model card says “A ~6.2B-parameter backbone with modality-specific pathways (text / vision / generation)”. The paper says “the base RxBrain model contains 6.2 B parameters”.
- Base model
- HY-Embodied-0.5Source and note: Base modelB arxiv.org/html/2607.14187 · Retrieved 9 November 2026Note. The paper says “Built upon HY-Embodied-0.5”.
- Licence type
- Free, commercial useSource and note: Licence typeC huggingface.co/tencent/Hy-Embodied-RxBrain-1.0/blob/main/LICENSE · Retrieved 9 November 2026Note. Licence file: “Terms of the Apache-2.0:” covering the code and model parameter weights (“Hy-Embodied-RxBrain-1.0 refers to the code and model parameter weights made publicly available by Tencent.”)
Specifications
8 of 14 stated 6 not stated
3 fields hidden 3 not-stated fields with a note stay ·
How to read the figuresShow guideClose
- not statedNo source
- 0Reported as zero
- ASource
Select a source marker for source details and notes.
Release1/1 disclosed
Released15 July 2026
Source and note: ReleasedD arxiv.org/abs/2607.14187 · Retrieved 9 November 2026Note. [A] “[Submitted on 15 Jul 2026]”
Input and output2/4 disclosed
Model typeEmbodied reasoning
Source and note: Model typeB arxiv.org/html/2607.14187 · Retrieved 9 November 2026Note. The paper describes “a more unified architecture that enables embodied reasoning, world state prediction, and joint language-visual subgoal planning”. It sets the model apart from both vision-language models and generative world models: “Unlike vision-language models that emphasize scene understanding and textual decision making, or generative world models that mainly predict future visual states”.
InputsCamera Language
Source and note: InputsA huggingface.co/tencent/Hy-Embodied-RxBrain-1.0 · Retrieved 9 November 2026Note. The model card lists “question answering and chain-of-thought over images and multi-frame video”.
What it controlsNot stated
Source and note: What it controlsA huggingface.co/tencent/Hy-Embodied-RxBrain-1.0 · Retrieved 9 November 2026Note. The model card describes text and imagined-frame output: “alternates reasoning text and flow-matched imagined frames”. The paper describes robot action generation as an extension: “We also extend RxBrain to continuous robot action generation, where it shows promising real-robot performance without large-scale action-data pretraining.”
Degrees of freedom controlledNot stated
Size and speed1/2 disclosed
Parametersapproximately 6.2B
Source and note: ParametersA huggingface.co/tencent/Hy-Embodied-RxBrain-1.0 · Retrieved 9 November 2026Note. The model card says “A ~6.2B-parameter backbone with modality-specific pathways (text / vision / generation)”. The paper says “the base RxBrain model contains 6.2 B parameters”.
Control rateNot stated
Where it runs and what it is built on1/2 disclosed
On robot or cloudNot stated
Base modelHY-Embodied-0.5
Source and note: Base modelB arxiv.org/html/2607.14187 · Retrieved 9 November 2026Note. The paper says “Built upon HY-Embodied-0.5”.
Openness3/3 disclosed
AccessOpen weights (download)
Source and note: AccessA huggingface.co/tencent/Hy-Embodied-RxBrain-1.0 · Retrieved 9 November 2026Note. The model card says Tencent released “the technical report, official inference code, and model weights.”
Licence documentApache-2.0
Source and note: Licence documentC huggingface.co/tencent/Hy-Embodied-RxBrain-1.0/blob/main/LICENSE · Retrieved 9 November 2026Note. Model card: “license: apache-2.0”. Licence file: “Hy-Embodied-RxBrain-1.0 refers to the code and model parameter weights made publicly available by Tencent.”
Licence typeFree, commercial use
Source and note: Licence typeC huggingface.co/tencent/Hy-Embodied-RxBrain-1.0/blob/main/LICENSE · Retrieved 9 November 2026Note. Licence file: “Terms of the Apache-2.0:” covering the code and model parameter weights (“Hy-Embodied-RxBrain-1.0 refers to the code and model parameter weights made publicly available by Tencent.”)
The developer’s own claims0/2 disclosed
Learns new tasksNot stated
Source and note: Learns new tasksB arxiv.org/html/2607.14187 · Retrieved 9 November 2026Note. The paper's limitations section says “fully out-of-distribution environments, embodiments, or task domains still require additional fine-tuning to achieve better performance”.
Works across robotsNot stated
Source and note: Works across robotsB arxiv.org/html/2607.14187 · Retrieved 9 November 2026Note. The paper tests its robot action extension on two robots: “We further evaluate our method on two real-world robot embodiments: the DOBOT X-Trainer and an ARX dual-arm platform”.Yes only if the developer itself says the model runs on different robot models, for example robots from two manufacturers or two of one manufacturer’s models. Several units of the same robot do not count.
Notes
Tencent Robotics X, Futian Laboratory and the Tencent Hy Team made RxBrain.
Tencent describes RxBrain as an embodied cognition foundation model with joint language-visual reasoning and imagination.
The model card lists three capabilities: embodied understanding and reasoning, world state prediction and joint subgoal planning.
Tencent released the model weights and inference code in July 2026.
Sources
One line per source, not per figure: a source has one retrieval date.
- Ahuggingface.co/tencent/Hy-Embodied-RxBrain-1.0Main source. Retrieved 9 November 2026
- Barxiv.org/html/2607.14187Retrieved 9 November 2026
- Chuggingface.co/tencent/Hy-Embodied-RxBrain-1.0/blob/main/LICENSERetrieved 9 November 2026
- Darxiv.org/abs/2607.14187Retrieved 9 November 2026
More models
Same developer
- No photograph
HY-Embodied-0.5Tencent Robotics X
Hy-Embodied-0.5-VLATencent Robotics X
- No photograph
Hy-Embodied-VLM-1.0Tencent Robotics X
