As emphasized by the company, these servers are prepared for local work with generative models, in particular — with DeepSeek. "Graviton" explained:
The choice of DeepSeek is due to the openness, scalability, and readiness of the AI model for practical use. On its basis, it is possible to deploy services with reliable protection of key data, configure systems and work scenarios taking into account the specifics of the industry — industry, healthcare, logistics, banking, etc. Despite the overall complexity of the technological stack and the rapid updating of AI models, the "Graviton" server solution with DeepSeek is optimized and ready for high-performance work.
The DeepSeek solution supports AI operation for the x86 architecture, launching PaaS services for creating software with AI. It includes working with a wide range of model repositories and deploying services "in one click".
Among the main technical characteristics are up to 3 dual-slot GPUs or 6 single-slot GPUs, up to 2 lntel Xeon Scalable 4th and 5th generation processors, up to 8 TB of RAM, up to 27 SFF HDDs with hot-swap, a total of up to 29 modules, a 2U high case, compatibility with RED OS and ROSA "Хром".