Integration
Secure computation, sealed in hardware
Inference leading LLMs run inside verifiably secure runtimes, powered by Intel TDX and NVIDIA Confidential Computing architectures.
1import OpenAI from "openai"
2
3const zima = new OpenAI({
4 baseURL: "https://zimalabs.io/api/v1",
5 apiKey: process.env.ZIMA_KEY,
6})
7
8const response = await zima.chat.completions.create({
9 model: "qwen3-next-80B-a3B-instruct",
10 messages: [{ role: "user", content: "..." }],
11})
12
13// encrypted in hardware. zero data retained.Private AI for whatever you can’t afford to leak
The same Zima, whether you’re asking a personal question or building on the API.
Everyday
Ask about health, money, or anything personal. Your words are never stored or used to train a model.
Sensitive work
Legal, health, and finance teams that can't leak client data. Privilege and confidentiality stay intact.
Developers
Ship apps where open-source prompts run sealed in a hardware enclave, never logged or retained. One OpenAI-compatible key.
From SDK request to protected output, without storing prompts or responses.
Verified inference path
Chat privately. Sealed in hardware.
Open-source models run sealed inside secure hardware, isolated in memory from the host. What you type stays yours. Never logged, never sold, never turned into training data.
Sealed in the enclave
Your messages run inside a hardware enclave, sealed in memory and isolated from the host the whole time.
Attach files privately
Add a file to the conversation. It's processed inside the enclave and never stored.
Never trained on
Your conversations are never used to train a model. Not ours, not anyone's.
| Model | Provider | Input / 1M | Output / 1M |
|---|---|---|---|
| GPT OSS 20B | OpenAI | $0.020 | $0.100 |
| GPT-OSS 120B | OpenAI | $0.100 | $0.400 |
| Qwen3 Next 80B Instruct | Alibaba Cloud | $0.200 | $1.600 |
| DeepSeek V3 | DeepSeek | $0.270 | $1.100 |
| Kimi K2.5 | Moonshot | $0.650 | $3.100 |
Explore the models
The open models you can run, with transparent per-token pricing and zero intermediary markup.
Sealed so the host can’t see it
Chat and API both run locked inside the same sealed hardware. Unreadable to anyone outside it, including Zima, and kept nowhere.
Your data is yours. Your chats stay local. Zima never has your data, and will never sell it.
Secure chat
Chat privately. Nothing you say is stored.
Open chat→Secure API
Same privacy for what you build.
Create a key→Seamless migration to secure compute
Use your existing OpenAI client. Swap the base URL and API key to make every inference confidential today.
Install
Drop in the OpenAI SDK. No proprietary client, no vendor lock. One dependency you already know.
Point to Zima
Swap the base URL. Open-source models run inside our secure enclave — sealed in hardware. Every request: zero data retained.
Ship to Production
Same models, same latency, hardware-level encryption. Your compliance team will thank you.
Frequently asked questions
Security, compatibility, and pricing.
No. Requests exist only for the duration of execution inside the enclave. No prompts, outputs, metadata, or logs are persisted. This is enforced by hardware, not policy.
Point the OpenAI SDK at Zima, keep model choice and pricing visibility, and move sensitive traffic into attested hardware.