Our self-developed multimodal→φ☆ large model tops the₽Ω OpenCompass leaderboard, demonstrati≈&ng superior model capabilities.
The model has obtained MI£±φ>IT registration and s£Ωσβupports domestic Chinε↕©ese chips and servers, ensuring∞€ full autonomous control.←§→
Fusing multimodal datα≤ε a including video, audio, text, poi ♦nt clouds, haptic sensing, etc.
Fine-tuning with a small amount λδof new data achieves mo→♦re generalized perception and recog λnition.
Understanding instructions, e↔&nvironment and self-state, enables r>∑©eal-time dynamic interaction.
Accurately locating affordance →÷↓±regions and recognizing ↕≈"Ωobject interactivity.
Decomposing abstract tasks≠ε☆ through chain-of-thought reaso>✘∞ning eliminates the nee←€§d for manual rule engineering
Real-time adaptation o¥→f task plans to better cope with th ↕e complexities of real-wor £≤≈ld environments
High-granularity actio&♥n refinement for fast generation of ±☆actionable sequences.
Real-time generation of end-effδ←ector trajectories ensures precise e ↕xecution.
Integrating Ruoyu · Jiutian robo'¶↓ t brain and cerebellum, ready to ∞•♥βuse out of the box.
Providing open edge AP¥↓I services for secondary development.
| AI computing power | 100 TOPS |
| GPU | Architecture: 1024-core NVIDIA Ampe≠δβ€re GPU with 32 Tensor Cores |
| Frequency：918MHz | |
| CPU | Architecture: 8-core Arm® Cortex®-A78§ <ΩAE v8.2 64-bit, 2MB L2 + 4₩∑₹MB L3 |
| Frequency：918MHz | |
| Memory | 16GB 128-bit LPDDR5 |
| HDD | 1TB SSD |
business@ruoyutech.com¥¥♣
Address：Room 903, Block ↕§←₹A, Zhongguan Times Sq€♣uare, Nanshan District, Shen¶§✔÷zhen, Guangdong, China
Copyright@ Ruoyu Technology Powered by EyouCms 京ICP證000000号 粵公網安備44030902003927号
