Mifeng jointly launches the embodied evaluation platform RoboColiseum, with GPT-6, Opus 5.5, and Gemini 4 competing head-to-head
Mifeng Technology, together with nearly 20 institutions including Hugging Face, NVIDIA, and Tsinghua University, jointly launched the embodied evaluation platform RoboColiseum, and announced this round's results on October 20: Claude Opus 5.5 ranked first overall among the four flagship models, followed closely by GPT-6 Astra. GPT-6 Astra scored 99.35 on instruction following, while Opus 5.5 performed better on spatial understanding and general manipulation and consumed the fewest tokens. The platform contains 78 simulated evaluation tasks, with a simulation-to-real-world correlation of 89.5%, and since launching in August, more than 100 teams have participated.
Body · AI translation
Mifeng co-initiates embodied evaluation platform RoboColiseum, where GPT-6, Opus 5.5, and Gemini 4 face off on the same stage! Zhong Lilei 1 hour ago. Building a public, neutral, and reproducible "arena" and "training ground" for embodied intelligence models.
When the flagship models of the three AI giants OpenAI, Anthropic, and Google face off on the same stage, who is better at controlling robots?
On October 20, RoboColiseum, an embodied arena jointly initiated by nearly 20 universities, enterprises, and research institutions worldwide including Mifeng Technology, Hugging Face, NVIDIA, and Tsinghua University, announced the evaluation results for this round: Claude Opus 5.5 ranked first overall among the four flagship models, with GPT-6 Astra close behind.
Breaking it down by category, among the four flagship models, GPT-6 Astra leads in instruction following and overall perturbation adaptation, with an instruction-following score of 99.35; Opus performs better in spatial understanding and general manipulation, and took first place overall in the flagship group with the lowest token consumption across all evaluation rounds. In tests with perturbed image quality, GPT-6.1 Sol lagged noticeably behind the other three models, revealing differences in stability among models when responding to changes in visual information.
However, the top two spots in this round's overall ranking are still occupied by embodied policy models, which maintain an advantage especially in the general manipulation dimension.
This head-to-head contest provides an intuitive reference for the industry to see the strengths and weaknesses of general large models and embodied policy models.
RoboColiseum focuses on instruction following, spatial understanding, perturbation adaptation, and general manipulation, and has launched four capability sub-leaderboards and 78 high-fidelity simulation evaluation tasks. It presents both the per-category performance of models and records the task completion steps and failure locations, helping developers quickly pinpoint model capability weaknesses.
As a co-initiator, Mifeng provides RoboColiseum with high-fidelity object assets and simulation scenarios, supports the construction of evaluation environments for complex robotic manipulation tasks, and provides foundational resources for testing model capabilities under different object interactions, spatial layouts, and environmental conditions.
Whether a model's results in simulation can reflect real-world manipulation performance is also key to evaluation. According to the actual test results published by the platform, the correlation between RoboColiseum simulation evaluation and real-machine deployment reaches 89.5%, and the evaluation difference for the same model between simulation and the real world is less than 10%, which can provide a reference for developers in model screening and problem localization before entering real-machine testing.
Mifeng Technology provides a data supply system that is "comprehensive, high-quality, and fast," covering three major data paradigms: real-machine, simulation, and embodiment-free. In simulation, Mifeng provides two types of resources, object assets and environment assets, covering a variety of everyday objects and indoor scenes, and offers capabilities such as scene reconstruction, expert trajectory data acquisition, and visualization of evaluation results, helping model teams carry out diverse training and evaluation.
Coliseum originally refers to the ancient Roman circular arena. RoboColiseum uses this as its imagery to build a multi-dimensional systematic evaluation system, and with simulation evaluation highly consistent with real-machine performance, to build a public, neutral, and reproducible "arena" and "training ground" for embodied intelligence models. Any university, research institution, or enterprise developer can submit their model to undergo a standardized "exam." Since its launch in August, more than 100 teams at home and abroad have carried out model training and evaluation on the platform.
RoboColiseum is now open to global developers, and model teams are welcome to participate in evaluation.
Source: Submission
The images in this article are from: 123RF licensed image library. Infringement report. Back. Share. Related articles. All comments.