Translated with AI from the Russian original.
The platform supports popular models including Llama, Qwen, Mistral, Mixtral, GPT-OSS, DeepSeek, Phi, and Gemma.
Trusta, Adata Technology's enterprise storage brand, has introduced the Trusta AI Scaler Extended Memory Solution – a new architecture for local enterprise AI deployment. The solution allows for the expansion of AI model deployment beyond the GPU to system memory and high-speed SSDs, reducing implementation costs by more than 50%.
As explained by the company, the Trusta AI Scaler Toolkit is a comprehensive hardware and software solution that helps enterprises overcome GPU memory limitations and optimize resources. In inference and fine-tuning scenarios, the solution dynamically allocates computing resources between the GPU, DRAM, and SSD.
The platform supports popular models including Llama, Qwen, Mistral, Mixtral, GPT-OSS, DeepSeek, Phi, and Gemma, as well as AI agent applications. The AI Scaler Toolkit is now available for download and use.
