Case Study: Aion Labs achieves faster, more accurate LLM performance with SambaNova Cloud

Aion Labs uses SambaCloud to improve LLM performance

Aion Labs, an AI model company, needed to speed up and improve the accuracy of its multi-model LLM workflows, which involve many sequential calls and can create unacceptable latency for interactive and agentic AI use cases. To support this, Aion Labs turned to SambaNova Systems and its SambaCloud service to enable high-speed inference on large open-source models.

Using SambaNova Systems’ SambaCloud, powered by the SambaNova RDU SN40L, Aion Labs combined models such as DeepSeek R1 671B, distill variants, Llama models, and its own fine-tuned models into a more accurate system. The result was much faster first-pass performance—SambaNova delivers the right outcome the first time 80%-90% of the time versus 15% for most models—helping Aion Labs serve more than 10 million monthly active users and complete tasks more quickly.

Aion Labs

Kacper Madej

Chief Technology Officer