Private Inference API
Simple. Secure. Automatically verified.
Interact with Tinfoil using the same OpenAI Chat Completions API standard:
- from openai import OpenAI
+ from tinfoil import OpenAI
Tinfoil client
The Tinfoil client provides secure, verified access to our API: Requests are automatically blocked if verification fails. For existing OpenAI integrations, Tinfoil offers a compatible API endpoint connected to a secure enclave making it a drop-in replacement.
$ tinfoil http post "https://inference.delta.tinfoil.sh/api/chat" \
-e inference.delta.tinfoil.sh \
-r tinfoilanalytics/provably-private-deepseek-r1 \
-b '{
"model": "deepseek-r1:70b",
"messages": [
{
"role": "system",
"content": "You are a helpful assistant."
},
{
"role": "user",
"content": "Why is tinfoil now called aluminum foil?"
}
]
}'
Installation:
# Download binary and checksums
curl -LO https://github.com/tinfoilanalytics/tinfoil-cli/releases/download/v0.0.2/tinfoil-cli_0.0.2_darwin_arm64.tar.gz
curl -LO https://github.com/tinfoilanalytics/tinfoil-cli/releases/download/v0.0.2/tinfoil-cli_0.0.2_checksums.txt
# Verify checksum (grep for relevant architecture only)
grep darwin_arm64 tinfoil-cli_0.0.2_checksums.txt | shasum -a 256 -c
# Extract binary
tar xf tinfoil-cli_0.0.2_darwin_arm64.tar.gz
# Move to /usr/local/bin
sudo mv tinfoil /usr/local/bin/
Available Models
DeepSeek-R1-Distill-Llama-70B
High-performance reasoning model with exceptional benchmarks
Model ID:
deepseek-70b
Llama 3.2 1B
General purpose language model optimized for efficiency
Model ID:
llama3.2-1b
Llama Guard 3 1B
Safety-focused model for content filtering and moderation
Model ID:
llama-guard3-1b
Qwen 2.5 Coder 0.5B
Specialized model for code generation and analysis
Model ID:
qwen2.5-coder-0.5b
Nomic Embed Text
Fast and efficient text embedding model
Model ID:
nomic-embed-text
More models coming soon!
Join the private preview for early access