あなたのコーディングエージェントはコードベース全体をスキャンできるため、その情報を安全に保つことが不可欠です。このコンテキストがいずれかの段階であなたのコンピュータから外部へ出る前に、Private AI Proxy がAIサービスを認証します。ソフトウェアは、フェイルクローズのフォワーディング、ハードウェアアテステーション、TLSキー・ピニングを実装することで、この保護を保証します。
Over the last 24 hours, Phala Confidential AI processed a remarkable 78.09 billion tokens. This secure platform operates entirely on TEE-backed AI infrastructure, providing a reliable foundation for all user interactions.
During this timeframe, three specific choices stood out among our most utilized models. GLM-5.3 took the top spot, making up 43.54% of the total volume. Following in second place was GLM-5.3 Flash with a 26.96% share, while DeepSeek V4 Flash captured 11.46% of the network activity.
For further details, you are welcome to visit https://phala.com/models/z-ai/glm-5.3
We are pleased to share that Qwen3 8B is now officially accessible on Phala. The model is securely hosted within a TDX-attested GPU TEE to deliver completely private and verifiable inference. This reliable setup makes it an excellent choice for a variety of advanced tasks, including reasoning, coding, and the deployment of long-horizon agents. You can explore and utilize the model today by visiting https://phala.com/models/qwen/qwen3-8b