Confidential AI Playground — Sovereign Models in Intel TDX

Test Qwen3-235B-TEE, DeepSeek-R1-TEE, Qwen3-32B-TEE without signup

VoltageGPU Playground runs three open-weight LLMs sealed inside Intel TDX hardware enclaves. Anonymous visitors get 3 free messages with no signup, no credit card. The underlying API at api.voltagegpu.com/v1 is OpenAI-compatible — point your existing OpenAI SDK at the new base URL and your prompts run in EU-based sovereign infrastructure.

Available TEE models

• Qwen3-32B-TEE — 40K context, $0.15/M input tokens, fast general reasoning
• Qwen3-235B-A22B-Instruct-2507-TEE — 262K context, flagship, multilingual
• DeepSeek-R1-0528-TEE — 163K context, hardened reasoning model

BYOA — Bring Your Own Agent

The playground is the lightweight surface. For production, point CrewAI, LangChain, OpenAI SDK or any MCP client at the same OpenAI-compatible endpoint. Your agent code runs wherever you want; inference runs in our TEE. See BYOA confidential AI.

Why a confidential playground matters

Public AI playgrounds (OpenAI, Google AI Studio, Anthropic console) log every prompt, train on it by default, and ship it through US infrastructure. VoltageGPU Playground seals every request inside Intel TDX, retains nothing, and runs in EU.

Related

• Confidential AI inference platform
• OpenAI-compatible API reference
• All TEE models with pricing
• Attestation report