A Trustless, Private Compute Layer

One API, all of AI.

Verified runs. Privacy on demand. Zero platform fees.

Get API Key Explore models

Requests

Tokens

TEE Proofs

Models

Providers

Requests

Tokens

TEE Proofs

Models

Providers

Quick Start

Install Claude Code

ANTHROPIC_AUTH_TOKEN='<YOUR_API_KEY>' claude --settings '{"env":{"ANTHROPIC_BASE_URL":"https://router-api.0g.ai","ANTHROPIC_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_FABLE_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_OPUS_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_HAIKU_MODEL":"0gm-1.0-35b-a3b","CLAUDE_CODE_MAX_CONTEXT_TOKENS":"1105920","CLAUDE_CODE_ATTRIBUTION_HEADER":"0","ENABLE_TOOL_SEARCH":"true","CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS":"1"},"fallbackModel":["0gm-1.0-35b-a3b"],"modelOverrides":{"claude-sonnet-5":"0gm-1.0-35b-a3b"}}'

Run from project folder. macOS/Linux; Windows: WSL/Git Bash. No files saved or Claude settings changed. glm-5.3 (edit JSON).

Install Hermes Agent

ANTHROPIC_AUTH_TOKEN='<YOUR_API_KEY>' claude --settings '{"env":{"ANTHROPIC_BASE_URL":"https://router-api.0g.ai","ANTHROPIC_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_FABLE_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_OPUS_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_HAIKU_MODEL":"0gm-1.0-35b-a3b","CLAUDE_CODE_MAX_CONTEXT_TOKENS":"1105920","CLAUDE_CODE_ATTRIBUTION_HEADER":"0","ENABLE_TOOL_SEARCH":"true","CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS":"1"},"fallbackModel":["0gm-1.0-35b-a3b"],"modelOverrides":{"claude-sonnet-5":"0gm-1.0-35b-a3b"}}'

Run from project folder. macOS/Linux; Windows: WSL/Git Bash. Config/API key in temp folder only. glm-5.3 (edit command).

Install OpenCode

ANTHROPIC_AUTH_TOKEN='<YOUR_API_KEY>' claude --settings '{"env":{"ANTHROPIC_BASE_URL":"https://router-api.0g.ai","ANTHROPIC_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_FABLE_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_OPUS_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_HAIKU_MODEL":"0gm-1.0-35b-a3b","CLAUDE_CODE_MAX_CONTEXT_TOKENS":"1105920","CLAUDE_CODE_ATTRIBUTION_HEADER":"0","ENABLE_TOOL_SEARCH":"true","CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS":"1"},"fallbackModel":["0gm-1.0-35b-a3b"],"modelOverrides":{"claude-sonnet-5":"0gm-1.0-35b-a3b"}}'

Run from project folder. macOS/Linux; Windows: WSL/Git Bash. 0G Router for this run only; no files written. Model: glm-5.3.

Install pi

ANTHROPIC_AUTH_TOKEN='<YOUR_API_KEY>' claude --settings '{"env":{"ANTHROPIC_BASE_URL":"https://router-api.0g.ai","ANTHROPIC_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_FABLE_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_OPUS_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_HAIKU_MODEL":"0gm-1.0-35b-a3b","CLAUDE_CODE_MAX_CONTEXT_TOKENS":"1105920","CLAUDE_CODE_ATTRIBUTION_HEADER":"0","ENABLE_TOOL_SEARCH":"true","CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS":"1"},"fallbackModel":["0gm-1.0-35b-a3b"],"modelOverrides":{"claude-sonnet-5":"0gm-1.0-35b-a3b"}}'

Run from project folder. macOS/Linux; Windows: WSL/Git Bash. Config/API key in temp folder only. glm-5.3 (edit config and --model).

API documentation

ANTHROPIC_AUTH_TOKEN='<YOUR_API_KEY>' claude --settings '{"env":{"ANTHROPIC_BASE_URL":"https://router-api.0g.ai","ANTHROPIC_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_FABLE_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_OPUS_MODEL":"glm-5.3","ANTHROPIC_DEFAULT_HAIKU_MODEL":"0gm-1.0-35b-a3b","CLAUDE_CODE_MAX_CONTEXT_TOKENS":"1105920","CLAUDE_CODE_ATTRIBUTION_HEADER":"0","ENABLE_TOOL_SEARCH":"true","CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS":"1"},"fallbackModel":["0gm-1.0-35b-a3b"],"modelOverrides":{"claude-sonnet-5":"0gm-1.0-35b-a3b"}}'

API key included. No setup form.

Try in Playground

Loading preview… Eight cinematic portraits in warm light, each with a surreal twist, generated with Z-Image-Turbo on 0G

Image

Trustless Compute Layer

How 0G protects your AI requests

Trusted privacy

Encrypted end to end through the router; sensitive workloads run inside hardware enclaves. Not even 0G can read your prompts — and attestation lets you check, not believe.

Private

Trusted routing

You choose which providers to trust. Every request is verifiably routed to the channel you picked — no silent substitutions, no cheap-model swaps — with a signed receipt to prove it.

Verified

Trusted models

The model you call is the model you get. TeeML proves it with hardware attestation, TeeTLS proves the official endpoint, and SPML proves it with post-hoc sampling.

Verified

Marketplace

Live models, real pricing

Premium, verifiable models — while staying remarkably competitive on price.

View all models

ByteDance Seedance 2.5

by 0G Foundation

per 1M tokens

In Out $8.4240

MiniMax-H3

by 0G Foundation

per 1M tokens

In Out $0.1950

GLM-5.3

by 0G Foundation

1M ctxtools

per 1M tokens

In $1.4000Out $4.4000

View all models

App Leaderboard

Popular Apps

0G App

0G App is your gateway to decentralized, verifiable AI—chat with models, build apps, and automate workflows in your browser.

AI · DEV TOOLS

Try

Neosoul

Neosoul makes AI agents measurable, ownable, aligned, controllable, and accountable.

AI AGENT · PRODUCTIVITY

Try

Claraverse

ClaraVerse is a private, open-source AI workspace to chat, build agent teams, and automate workflows on your infrastructure.

AI WORKSPACE · AUTOMATION

Try

Trust Mode

Not every task needs the same privacy.

Widest coverage → Strongest guarantee

Standard Routing

Every model and provider on the network, including community-hosted channels — best price and availability across the full catalog.

Best forEveryday tasks

Everything in Verified, plus community channels

Verified

Verified Routing

Routes only to endpoints with verified provenance, so you always know which model ran. Privacy depends on the endpoint handling the request.

Best forApps that need provenance

Everything in Private, plus official providers

PrivateTEE · E2EE · ZDR

Private Inference

Combines hardware-isolated inference, end-to-end encrypted routing, and zero data retention, while remaining cryptographically verifiable — no one can see your data, not even us.

Best forSensitive data

Endpoints with TEE isolation, end-to-end encryption, and zero data retention

Encrypted Transit

Encrypted to the broker. Graded beyond.

Encrypted routing is rolling out per model — see the model list.

End-to-end encrypted TEE routing diagram

Set Trust Mode

Switching costs one line.

A header for a single call, or a default on your API key — no new endpoint, no SDK changes.

0G Dashboard — API Keys

Key name (e.g. production)

+ Create

Trust mode

StandardAll providers

VerifiedVerifiable only

PrivateTEE · E2EE · ZDR

Standard routing

Per request — add a header

Set this header on any call. Omit it and the key's default applies.

X-0G-Provider-Trust-Mode:verified|private

Per key — set a default

Choose a trust mode when you create an API key. Every request it makes uses that mode unless a header overrides it.

Verifiability

Don't trust. Verify.

Verification and privacy are two different promises. Verification proves the right model ran, untampered. Privacy is about who can read your data along the way.

0G reports both, honestly, for every request: the path your data takes, hop by hop — and exactly who could see it. No fine print.

Path Verifier

Gateway

Your request is encrypted inside the gateway enclave. No one can see your data.

Attested

Router

Forwards ciphertext and settles billing. Nothing to decrypt, nothing to trust.

Ciphertext only

Broker

Your request is decrypted inside the provider's broker enclave — not on an open server.

Attested

Inference

The model runs in a hardware-isolated environment, protected from the provider and 0G.

PrivateTEE · E2EE · ZDR

Privacy

PrivateTEE · E2EE · ZDREncrypted Transit and isolated inference keep prompts unreadable to the provider and 0G.

Execution

VerifiedThe model and execution environment are cryptographically verified.

Private path verified.

View

FAQ

Common questions

Official models: 0G verifies that requests reach an authoritative, officially-backed model channel.

Private endpoints: Private requires hardware-isolated inference, end-to-end encrypted routing, and zero data retention.

Other open models: Open-model deployments can provide cryptographic evidence of the model and hardware that ran.

Verified endpoints return cryptographic evidence that the response came from the expected model and execution environment. Private endpoints also provide hardware-isolated inference, end-to-end encrypted routing, and zero data retention.

Standard is the default and prioritizes full catalog coverage, price, and availability. Choose verified when you need model and route provenance. Choose private when the request requires TEE isolation, end-to-end encryption, and zero data retention.

Option 1 — Request header. Set the trust mode on a single request:
X-0G-Provider-Trust-Mode: verified | private
Use verified for verifiable endpoints, or private for endpoints with TEE isolation, end-to-end encryption, and zero data retention. Omit the header for Standard.

Option 2 — API key default. Set a key's trust mode to Verified or Private in the dashboard, and every request it makes uses that mode unless a header overrides it.

Path: Dashboard → API Key → ⋯ → Edit

Private Computer supports both 0G Mode and USD Mode. In 0G Mode, deposits, balances, and usage charges are denominated in 0G. In USD Mode, you can pay with fiat through Stripe, with your balance and usage settled in USD.

Because the network is decentralized, multiple providers serve the same models. If one becomes unavailable, your request is automatically routed to another healthy provider serving that model. Your funds stay safe in your on-chain ledger throughout.

Providers set their own per-token rates, and the marketplace creates natural competition that drives prices down. You can compare prices across providers for the same model and route to the best value — frontier models come discounted, and open models go even lower.

It depends on your trust mode.

Private: requests run only on GPUs with hardware isolation (TEE), whether operated by 0G or third-party providers. The operator can't read your prompt, and attestation lets you check.

Verified: requests can also run on official model APIs, such as Alibaba, Zhipu, or MiniMax. 0G verifies the route, so you know which model answered. Privacy follows the provider's own policy.

Standard: everything above, plus community hosts and vetted resellers. Reseller endpoints are not disclosed.

Set the mode with the X-0G-Provider-Trust-Mode header, or as a default on your API key.

Buy compute once. Infinite AI

Stake 0G. Mint iAI. Unlock daily AI compute.

Get free credits