Telling people to verify their TEE sounds simple, but understanding what is genuinely being inspected is another matter. Our team has designed an interactive walkthrough to demystify the process. With this tool, you can follow individual attestation measurements, adjust your deployment inputs, and see the exact reasons behind any resulting shifts. We also provide published reference values for direct comparison.
Take a look at the guide here: https://cloud.phala.com/reference-values
Join Phala co-founder and CTO @bgmshana this Friday in San Francisco during the @sgl_project Summit 2026.
Hear the presentation, Production Private Inference Using SGLang and dstack, taking place on Nov 13 from 1:45–2:00 PM PT in the Decode Room at Fort Mason.
Explore further information at https://www.sglang.io/summit
Phala has officially launched support for FLUX.2 klein 4B, operating securely within a TDX-attested GPU TEE. This integration provides a confidential and verifiable inference solution designed for image generation tasks.
Access the deployment here: https://phala.com/models/black-forest-labs/flux.2-klein-4b
Operating on TEE-backed AI infrastructure, Phala Confidential AI successfully processed 120.50 billion tokens across the past 24 hours.
The leading models driving this volume include: 1. Kimi K3, accounting for 69.39% 2. Xing4.0 29B A4B, representing 10.71% 3. Qwen3.8 27B, contributing 8.61%
To explore the top-performing model, visit: https://phala.com/models/moonshotai/kimi-k3
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。