Private Inference

Building private inference for people and agents

GPT, Claude & more through one API key
Pay privately with stablecoins via @aztecnetwork using x402

[Private Inference @P_Inference]

🔓 Private Inference is now in Open Beta

No more whitelists - login now at https://t.co/HZDe589LmL

🧠 Frontier and open models side by side, in one chat and one OpenAI-compatible API. Fund a session once from your wallet. Each request is held at its maximum cost, and you pay only for what you actually use.

💸 Pay your way:
· USDC on Base
· cUSDC on Aztec, privately
· ZEC from any wallet
No model accounts, no cards, no monthly plans.

🔐 You decide how private each request is:
· Zero data retention, the default
· Standard routes, usually cheaper
· Private execution (beta): your browser encrypts the prompt to a model in an attested TEE and checks the attestation itself

🤝 Bring your friends. Your referral code sits at the top of every page, and friends you bring in earn you balance credit.

🚀 We ship every week: https://t.co/IZH04WjZlu

[Private Inference @P_Inference]

Pay for AI. Stay unseen. 🕶

Private Inference is one of the first apps on @aztecnetwork v6. From today at 9am ET, pay for GPT, Claude and open models from a private balance: what you spend, and from which wallet, stays off the public record.

What v6 brings:
⚡️ 1-second client-side proofs on native
🌉 Ethereum to Aztec deposits in as little as 12 seconds
🛡 3,000+ validators, no central operator to trust

Privacy shouldn't need permission.

https://t.co/DjJY2kE9wS

Private is better.

[Aztec @aztecnetwork]

Introducing Aztec Tierra (Alpha V6).

The platform for unstoppable private apps is here.

Ethereum's only decentralized L2, with fully programmable privacy: permissionless, private smart contracts.

No backdoors. No gatekeepers.

https://t.co/J1KZd6XT15

[Galactica.com @GalacticaNet]

Formal verification of the proving system and ~1s client-side proofs are the Tierra upgrades we care most about 🔐

Private inference @P_Inference only stays private if the payment rail does too. Glad to be building private AI on @aztecnetwork 🤝

[Private Inference @P_Inference]

UR (@urnetwork) is teaming up with @Persistent_AI on joint privacy research, here’s where where Private Inference fits in:

We’ll become a UR network operator, anonymising our traffic through UR's proxy network. Why? This lets us close the final privacy gap in our stack:

We cover payment privacy.
We cover prompt encryption & safeguarding.
No one covered network anonymity - until https://t.co/D7BvoSQ7kh showed up.

Soon https://t.co/D7BvoSQ7kh users will have the option of a PrivateInference-bundled subscription service too.

Say it with us: Private is Better

[URnetwork @urnetwork]

UR is partnering with @Persistent_AI. Through its product @P_Inference, Persistent AI is joining UR as our newest Network Operator.

Every AI request exposes three things about the person sending it:

1. Their identity
2. What they asked
3. Where they are connecting from

Private Inference has been building private AI that solves the first two. It gives people access to leading models such as GPT and Claude without storing their chats, and lets them pay privately.

UR solves the third. As a Network Operator, Persistent AI will route Private Inference's traffic over UR, where it is encrypted and sent out through connections shared by people globally, keeping users' IP addresses private.

Network Operators are where demand for SN25 alpha comes from. They pay for the traffic they use in alpha, so every new operator brings new demand to the subnet.

Both teams are passionate about AI and privacy, and the network is the first thing we are working on together.

Keep the internet private

[Private Inference @P_Inference]

Your coding agent can now run more privately: Private Inference added the Responses API, the protocol Codex and other agents speak. We don't store prompts or replies, and ZDR sends each request only to suppliers that keep nothing.

What that means for coders: one provider block in your Codex config, then codex --profile pi. Your agent keeps its tools and workflow, and you pick Standard or ZDR per request.

https://t.co/gZQQqrDGKc

[Private Inference @P_Inference]

🕶️ Pay for AI without revealing who you are. Prompt it without revealing what you asked.

[Private Inference @P_Inference]

What’s new since last week 🚀 (and something big on Friday 👀)

📎 Attach files to a message: notes, code, CSVs, logs (up to 10 files, 512 KB each). Drop them on the prompt or use the paperclip, and the model reads them with your message.

🔎 Web search in chat: tap the magnifier and the model gets up to ten pages from the web, cited right in the reply. Hover a citation to preview the page. Works with every model at the standard tier.

📚 Docs are live at https://t.co/OGRS1wDpke: who sees what at each privacy level, how private execution is checked, what we keep, payments and the API.

🚪 A new Start page: “Open app” takes you straight to sign-in, “Get an API key” drops you into your keys.

💬 Questions or feedback? There’s now a Discord link in the header.

🛡️ cUSDC on Aztec: if your cUSDC is sitting on your public balance, we now tell you that and how to move it to private, instead of “Balance too low.”

🧮 Math that looks like math: nested fractions and subscripts like d₁ now render properly.

And the big one 👇

Private Inference leaves closed beta and opens to everyone (Open Beta) on Friday 23 October. No invite needed.

One balance, every model, in chat or by API. Top up in USDC on Base, private cUSDC on Aztec or ZEC, and pay only for what you use.

Changelog: https://t.co/Smf0kssbOl

[Private Inference @P_Inference]

Private Inference keeps you private. End to End.

[Private Inference @P_Inference]

https://t.co/bhMkU2dh36

[Private Inference @P_Inference]

https://t.co/l790VDzFXv

[Private Inference @P_Inference]

https://t.co/pe9Ov5Q4km

[Private Inference @P_Inference]

We’re proud to be using @humntech bridged USDC (cUSDC) to power private payments on Private Inference. 🛡️

Most AI services want your card, your email and your payment history. With cUSDC on @aztecnetwork private rail, you can fund your account without exposing any of that on chain.

What this means for you:
→ 💰 Top up with cUSDC. The sender, recipient and amount stay off chain.

→ 🚫 No card, no monthly plan, no separate accounts with each model supplier.

→ 🔑 One balance and one API key give you access to GPT and Claude models.

→ 📊 Each request is held at its maximum cost, then charged only for actual usage. Unused funds are released right away.

A big thank you to the @humntech team. Their work on bridged USDC is what makes private stablecoin payments for AI possible.

[Private Inference @P_Inference]

What we’ve done since Oct 6 (closed-beta launch) 🚀

It’s been a busy week! Here’s what changed since the doors opened:

👀 Replies you can actually see: when a reply contains HTML or SVG, you now get the finished thing (a widget, a card, a chart) right in the chat, with the code one click away. It runs in a locked-down frame that can’t touch your account or chats.

⚡ No more freezing: long chats stay smooth while a reply streams in, and bold, code and lists format as they arrive instead of showing up as stray symbols.

🔑 API in under a minute: a new key now comes with a ready-to-run command (macOS, Linux or Windows) and a button that fires a test request.

💳 Fewer payment and sign-in headaches: cancelling a payment no longer throws an error, Aztec payments work again, and when sign-in fails you’re told what to do next instead of getting an error code.

🤝 Invite your friends: as of today, everyone on the whitelist can invite up to 10 people, and everyone who gets invited can invite up to 5 more. Your referral code and a copy-link button live at the top of every page.

🎨 Code that’s easier on the eyes: syntax highlighting in both dark and light themes.

✨ And a lot more: mobile menu, smarter top-ups, a handful of smaller fixes and polish all over the place.

Thanks to everyone testing and sending feedback, it’s shaping what we build next 💙

Stay posted in the Changelog: https://t.co/Smf0kssbOl

[Private Inference @P_Inference]

@pittpv 📨

[Private Inference @P_Inference]

https://t.co/Mtk0xFBEro is now live for private whitelisted beta phase 1.

What it is, why it exists, and how the whitelist works.

A thread 🧵

[Private Inference @P_Inference]

@aztecnetwork @GalacticaNet 🤝

[Aztec @aztecnetwork]

AI conversations hold some of the most sensitive data we create.

Private Inference tackles all three privacy layers: payment, identity, and words, with private payments live on Aztec today and private sessions coming with Alpha V6.

We're excited to announce @P_Inference as a V6 launch partner.

[Private Inference @P_Inference]

https://t.co/P19SBTq1bb

[Private Inference @P_Inference]

Two things give you away when you use AI:
the prompt and the payment.

Most private AI protects one of them, PI is built to protect both.

In private mode, your prompt is encrypted on your device and only an attested enclave can read it.

Payment settles privately when using Aztec, so nobody on-chain knows who paid.