# Unified Embodied AI: Pelican-Unified 1.0
_Pelican-Unified 1.0, the first unified embodied foundation model, achieves SOTA performance by integrating VLM, reasoning, and generation, proving unification enhances rather than compromises specialist strengths._
**Published:** 2026-05-15
**Source:** https://www.startuphub.ai/ai-news/ai-research/2026/unified-embodied-ai-pelican-unified-1-0
---
The pursuit of truly intelligent embodied agents has long been hampered by the need to train disparate, specialized models for perception, reasoning, and action. This fragmentation leads to inefficiencies and limits the holistic capabilities of AI systems. The introduction of [Pelican-Unified 1.0](https://arxiv.org/abs/2605.15153v1) marks a significant departure, presenting the first embodied foundation model built on the principle of unification.

Fragmented AI ModelsDriver

From the article 4 mentionsThe pursuit of truly intelligent embodied agents has long been hampered by the need to train disparate, specialized models for perception, reasoning, and action.Inefficiency & LimitsDriverFrom the articleThis fragmentation leads to inefficiencies and limits the holistic capabilities of AI systems.solvesPelican-Unified 1.0CoreFrom the article 3 mentionsThe introduction of Pelican-Unified 1.0 marks a significant departure, presenting the first embodied foundation model built on the principle of unification.usesUnified VLMCoresingle visual-language model maps diverse inputs to shared semantic spaceFrom the article 5 mentionsPelican-Unified 1.0 leverages a single Visual-Language Model (VLM) to serve as a unified understanding and reasoning module.enablesChain-of-Thought ReasoningCoreFrom the article 4 mentionsCrucially, it also performs autoregressive chain-of-thought reasoning, generating task- and action-oriented sequences in a single pass.Simultaneous OptimizationEffectFrom the articleThis unified approach allows for the backpropagation of language, video, and action losses into the shared representation, enabling simultaneous optimization of understanding, reasoning, imagination, and action, rather than relying on isolated expert systems.SOTA PerformanceOutcomeachieves state-of-the-art performance by integrating perception, reasoning, generationFrom the article 2 mentionsContrary to the intuition that unification might lead to diluted capabilities, Pelican-Unified 1.0 demonstrates that this paradigm can preserve and even enhance specialist performance.showsUnification EnhancesOutcomeproving unification enhances rather than compromises specialist strengthsFrom the article 2 mentionsContrary to the intuition that unification might lead to diluted capabilities, Pelican-Unified 1.0 demonstrates that this paradigm can preserve and even enhance specialist performance.

## Unifying Perception, Reasoning, and Imagination

Pelican-Unified 1.0 leverages a single Visual-Language Model (VLM) to serve as a unified understanding and reasoning module. This VLM maps diverse inputs, scenes, instructions, visual contexts, and [action](/ai-news/ai-research/2026/x-wam-bridging-action-and-4d-synthesis) histories, into a shared semantic space. Crucially, it also performs autoregressive chain-of-thought reasoning, generating task- and action-oriented sequences in a single pass. This unified approach allows for the backpropagation of language, video, and action losses into the shared representation, enabling simultaneous optimization of understanding, reasoning, imagination, and action, rather than relying on isolated expert systems.

## Specialist Strength Without Compromise

Contrary to the intuition that unification might lead to diluted capabilities, Pelican-Unified 1.0 demonstrates that this paradigm can preserve and even enhance specialist performance. A single checkpoint of the model achieved impressive results across multiple domains: 64.7 on eight VLM benchmarks (outperforming comparable-scale models), a first-place ranking of 66.03 on WorldArena, and 93.5 on RoboTwin (second-best among action methods). These findings underscore the efficacy of the unified approach in consolidating complex AI capabilities without sacrificing individual performance.
---
Original analysis from [startuphub.ai](https://www.startuphub.ai), the #1 AI startup directory.