Privatemode AI Reviews & Provider Details

Privatemode AI is a generative AI inference API developed by Edgeless Systems GmbH that provides access to open-source large language models with the technical guarantee that neither the vendor nor the infrastructure provider can access user prompts or model outputs. It offers an API interface compatible with both the OpenAI and Anthropic message formats. The privacy guarantee is enforced at the hardware level using AMD SEV-SNP confidential VMs and NVIDIA H100 GPUs running in confidential computing mode, hosted on Scaleway infrastructure. Encryption of prompts and responses is handled transparently by a local client-side proxy, which also performs remote attestation — a cryptographic verification step that confirms the server-side code matches published specifications before any data is transmitted. The source code is publicly available and builds are reproducible, meaning users can independently verify the deployment rather than relying on the vendor's assurances. Available models include gpt-oss-120b, Gemma 3 27B (with vision support), Qwen3-Coder 30B-A3B, Qwen3-Embedding 4B, Voxtral Mini 3B, Whisper large-v3, and Kimi K2.5. Supported API features include chat completions, embeddings, speech-to-text, tool calling, structured outputs, streaming, and prompt caching. Integrations include a desktop application for Windows, macOS, and Linux, and compatibility with coding tools such as Claude Code, VS Code extensions (Copilot, Cline, Continue), and JetBrains IDEs. Pricing has a free tier (5M initial tokens, then 1M tokens/month) and a pay-as-you-go standard tier with no monthly cap. Enterprise plans with custom rate limits are available.