Over the past 24 hours, the TEE-backed AI infrastructure of Phala Confidential AI has successfully processed a total of 81.33 billion tokens. An analysis of the most popular choices reveals that DeepSeek V4.1 Flash leads the group, accounting for 56.30% of the overall usage. The xing4.0-29b-a4b model follows with a 12.22% share, while qwen3.8-27b makes up 9.39% of the volume. You can find more information regarding our most utilized model by visiting https://phala.com/models/deepseek/deepseek-v4.1-flash.