去中心化推理Decentralised inference
Phala Network:机密 AI 云Phala Network: confidential AI cloud
交集:把私有模型、agent 与 GPU 任务放进硬件 TEE 里跑,并用证明报告替代「相信云厂商」。
适用范围:官方称支持 CPU TEE(Intel TDX)与 GPU TEE(NVIDIA CC),可把 Docker Compose 工作负载迁入 confidential VM;证明可被软件验证。
未解决:性能开销、GPU 型号覆盖、跨区域可用性与合规审计细节需逐项核对;TEE 不是模型正确性或输出质量的保证。Intersection: run private models, agents and GPU jobs inside hardware TEEs, and use attestation reports instead of trusting the cloud vendor.
Scope: the official page says it supports CPU TEE (Intel TDX) and GPU TEE (NVIDIA CC), can move Docker Compose workloads into confidential VMs, and exposes software-verifiable measurements.
Open: performance overhead, GPU model coverage, regional availability and audit details need case-by-case checks; TEE is not a guarantee of model correctness or output quality.
关键参数Key specs
- 协议状态 / status
- 已部署
- TEE / hardware
- Intel TDX / NVIDIA CC
- 部署形态 / deploy
- Confidential VM + Docker Compose
- 证明 / attestation
- 运行时度量与签名报告
官方原文怎么说What the source says
Run agents, private LLM models, and GPU jobs inside hardware-backed TEEs.
Move existing Docker Compose workloads into CPU or GPU confidential machines.
Phala emits runtime measurements that software can verify.
资料与许可Source and licence
- 资料时间Source checked
- 2026-11-19
- 许可 / 类型Licence / kind
- official-specification
- 许可说明Licence evidence
- https://phala.com/
- 类目Category
- 去中心化推理Decentralised inference