Tinfoil ⋅ Confidential AI

Private Inference API

Secure.Verified.

Interact with Tinfoil using the same OpenAI Chat Completions API standard:

-from openai import OpenAI
+from tinfoil import TinfoilAI

TinfoilAI supports the same interface as the OpenAI Python clientView on GitHub

Tinfoil client

The Tinfoil client provides secure, verified access to our API:

Requests are automatically blocked if verification of the enclavefails
Tinfoil is compatible with the OpenAI API endpoint, making it a drop-in replacement for existing OpenAI integrations

Model inference with Tinfoil Client

Select Model

High-performance reasoning model with exceptional benchmarks

Installation:

pip install tinfoil

Inference:

from tinfoil import TinfoilAI

client = TinfoilAI(
    enclave="deepseek-r1-70b-p.model.tinfoil.sh",
    repo="tinfoilsh/confidential-deepseek-r1-70b-prod",
    api_key="YOUR_API_KEY",
)

chat_completion = client.chat.completions.create(
    messages=[
        {
            "role": "user",
            "content": "Hello!",
        }
    ],
    model="deepseek-r1-70b",
)
print(chat_completion.choices[0].message.content)

Available Models

DeepSeek-R1-Distill-Llama-70B

High-performance reasoning model with exceptional benchmarks

Mistral-Small-3.1-24B

Advanced multimodal model with enhanced vision capabilities and extended context window

Llama 3.3 70B

High-performance multilingual language model for chat and reasoning

Llama 3.2 1B

Multilingual large language model optimized for dialogue

Llama Guard 3 1B

Safety-focused model for content filtering and moderation

Qwen 2.5 Coder 0.5B

Compact code-specialized model for lightweight applications

Nomic Embed Text

Open-source text embedding model that outperforms OpenAI models on key benchmarks