ᚠ · ᚢ · ᚱ · ᚾ · ᛊ · ᛃ · ᛟ · ᚨ · ᛞ · ᚲ · ᛏ · ᛒ · ᛗ · ᛚ · τ · ᚠ · ᚢ · ᚱ · ᚾ · ᛊ · ᛃ · ᛟ · ᚨ · ᛞ ᛊ · τ · ᚱ · ᛟ · ᚠ · ᛃ · ᚨ · ᛞ · ᚲ · ᛏ · ᛒ · ᛗ · ᛚ · ᛊ · τ · ᚱ · ᛟ · ᚠ ᚨ · ᛞ · τ · ᚱ · ᛊ · ᛟ · ᚠ · ᛃ · ᚲ · ᛏ · ᛒ · ᚨ · ᛞ · τ
τ

Open access to
decentralized AI

Access 1,397 open-source AI models through one OpenAI-compatible API. Drop-in replacement — change your base_url and every Llama, Mistral, DeepSeek, and Qwen model is available instantly. Decentralized inference across Bittensor's 128 subnet network. Free to start.

τ
1,397
AI Models
128
Subnets
<2s
Latency
Free
To Start

A unified API gateway to
Bittensor's decentralized AI network

OpenTau is a single endpoint that connects you to every AI model running on Bittensor's 128 subnets. Instead of managing connections to individual subnet miners, figuring out DRAIN micropayments, or running your own Bittensor node — you call one API.

It uses the same OpenAI SDK format you already know. No new libraries, no new syntax, no new concepts. Point your existing base_url to OpenTau and your code works with 1,397 decentralized models.

Under the hood, your requests route through Bittensor miners worldwide. No single company controls your AI access. No one can see your prompts. No one can cut you off.

Free tier: 1,000 req/day No signup friction No wallet required

τ
The Problem

AI is centralized.
That's the problem.

OpenAI, Google, Anthropic control your AI access. They can censor responses, rate-limit your app, price-gouge overnight, or cut you off entirely. Here's what that actually means for your product:

  • OpenAI can ban your account for policy violations — even accidental ones. Your app goes dark with no appeal.
  • Anthropic, Google, and OpenAI all have usage policies that restrict entire categories of applications. Build the wrong thing and you're out.
  • Rate limits change without notice. Your production app can break overnight because a provider decided to lower your tier.
  • Pricing increases are unilateral — you have no negotiating power. They raise prices, you pay or migrate.
  • Every prompt and response passes through their servers. You're trusting them with your IP, your users' data, and your competitive advantage.

Your entire product depends on a single provider. One policy change and your app breaks. One outage and your users get nothing. That's not infrastructure — that's a single point of failure you chose to depend on.

Meanwhile, Bittensor has 128 AI subnets with real capabilities — inference, training, search, storage, verification. But 125 of them generate zero revenue. $840M/year in emissions subsidize a network that only 3 subnets have figured out how to monetize. The infrastructure exists. The bridge to developers doesn't.

OpenTau is that bridge.

OpenTau routes through
128 subnets.

Bittensor's decentralized network means no single point of control. 1,397 models served by thousands of nodes worldwide. Censorship-resistant. Always available.

Here's how it works:

  1. 01Your request hits api.opentau.ai — same format as OpenAI.
  2. 02OpenTau authenticates your key and checks your credits.
  3. 03The request is forwarded to the best available miner on Bittensor's network.
  4. 04The miner runs the model and returns the response through OpenTau.
  5. 05Total latency: typically under 2 seconds. If one miner is slow, the request automatically routes to another.

You never interact with Bittensor directly — no wallets, no staking, no subnet mechanics. OpenTau handles everything. Change one line of code and your app runs on the most resilient AI infrastructure ever built.

τ

Why decentralized AI matters

No censorship

No single provider decides what you can or can't build. No content policies that shift overnight. No silent filtering of outputs. Your AI, unconstrained.

No single point of failure

The network stays up even when individual nodes go down. Thousands of miners across every continent ensure your requests always find a path to inference.

No vendor lock-in

Switch between 1,397 models freely. No proprietary formats, no migration costs, no "enterprise agreements." One standard API for the entire network.

Community-owned

Models trained by the community, for the community. Bittensor's incentive mechanism ensures the best models rise to the top — no corporate gatekeepers.

Change one line.
Access every model.

Same OpenAI SDK. Same request format. Same streaming. Just point your base_url to OpenTau and access 1,397 decentralized models instantly.

# change one line — that's it

from openai import OpenAI

client = OpenAI(
    base_url="https://api.opentau.ai/v1",
    api_key="your-opentau-key"
)

response = client.chat.completions.create(
    model="meta-llama/Llama-3.1-70B",
    messages=[{
        "role": "user",
        "content": "What is Bittensor?"
    }]
)
τ

1,397 Models

Llama 3.1, Mistral, DeepSeek R1, Qwen 2.5, Phi-3, Gemma 2 — every major open-source family, all through one endpoint. Switch between models by changing a string. Run the same prompt against ten different models to compare quality. No separate accounts, no per-provider billing.

Decentralized

Requests route through Bittensor's 128 subnets and thousands of miners worldwide. If one node goes down, your request automatically finds another path. No single company can revoke your access, throttle your usage, or read your prompts. Your AI infrastructure doesn't have an off switch.

Free to Start

1,000 requests per day on the free tier — no credit card, no wallet, no waitlist. Get an API key in seconds and start building immediately. When you're ready to scale, deposit USDC on Base. No subscriptions, no minimums, no surprise invoices.

OpenAI Compatible

Drop-in replacement for the OpenAI SDK. Same chat.completions.create() format, same streaming, same response schema. If your code works with OpenAI, it works with OpenTau. Migrate in 30 seconds — just change your base_url.

Usage Dashboard

Real-time analytics for every API call — requests, tokens, latency per model, cost breakdowns, error rates. See which models are fastest for your workload. Export usage data. Set spend alerts so you never get a surprise bill.

Bittensor Native

Built on the world's largest decentralized AI network — $1.5B ecosystem, 128 specialized subnets, 10,000+ miners globally. Not just a wrapper over someone else's GPU farm. Bittensor's incentive mechanism ensures the best models are always available and competing to serve your requests.

Every model. Decentralized.

Llama 3.1 8BLlama 3.1 70B Mistral 7BDeepSeek R1 Mixtral 8x7BQwen 2.5 72B Phi-3 MiniGemma 2 9B CodeLlama 34BYi 34B SOLAR 10.7BCommand R+ Nous Hermes 2WizardCoder Falcon 40BZephyr 7B + 1,379 more Llama 3.1 8BLlama 3.1 70B Mistral 7BDeepSeek R1 Mixtral 8x7BQwen 2.5 72B Phi-3 MiniGemma 2 9B CodeLlama 34BYi 34B SOLAR 10.7BCommand R+ Nous Hermes 2WizardCoder Falcon 40BZephyr 7B + 1,379 more

The full stack of decentralized AI

Not just chat completions. Six subnet-powered services, all behind one API key.

Inference

/v1/chat

SN58 + SN64

Fine-Tuning

/v1/fine-tuning

SN56 Gradients

Search

/v1/search

SN22 Desearch

Storage

/v1/storage

SN75 Hippius

Verification

/v1/verify

SN34 BitMind

Explore the stack
τ

Chain subnets. One API call.

Compose multi-subnet workflows in a single request. No centralized aggregator can replicate this.

Collect
SN13 Data
Train
SN56 Gradients
Serve
SN64 Chutes
See pipelines
τ

Built for builders who need sovereign AI

Crypto Developers

Building trading bots, DeFi agents, NFT tools, on-chain analytics. You need uncensored AI that won't refuse crypto-related prompts, crypto-native payments (USDC on Base), and zero KYC friction. OpenAI rejects half your use cases — OpenTau doesn't.

Agent Builders

Building autonomous AI agents that need inference, web search, file storage, and model fine-tuning. One API key for the full stack — not five different providers with five different billing systems. Works with ElizaOS, Virtuals, and any framework that speaks OpenAI format.

Privacy-First Teams

Healthcare, legal, finance, defense. You need inference where nobody can see your prompts — not even the API provider. OpenTau routes through decentralized miners with TEE-protected compute on the roadmap. No logs, no training on your data, no compliance risk from a third party reading your queries.

τ

Start free. Scale infinitely.

Free
$0
1,000 req/day
No deposit needed
Pay As You Go
~$0.0001/req
Deposit USDC on Base
No monthly commitment
Pro Credits
$50+ deposit
Fine-tuning + pipelines
Priority routing
Enterprise
Custom
Volume discounts
Confidential compute + SLA
Get started free
τ
ᚠ · ᚢ · ᚱ · ᚾ · ᛊ · ᛃ · ᛟ · ᚨ · ᛞ · ᚲ · ᛏ · ᛒ · ᛗ · ᛚ · τ · ᚠ · ᚢ · ᚱ · ᚾ · ᛊ · ᛃ · ᛟ · ᚨ · ᛞ · ᚲ · ᛏ · ᛒ · ᛗ ᛊ · τ · ᚱ · ᛟ · ᚠ · ᛃ · ᚨ · ᛞ · ᚲ · ᛏ · ᛒ · ᛗ · ᛚ · ᛊ · τ · ᚱ · ᛟ · ᚠ · ᛃ · ᚨ ᚨ · ᛞ · τ · ᚱ · ᛊ · ᛟ · ᚠ · ᛃ · ᚲ · ᛏ · ᛒ · ᚨ · ᛞ · τ · ᚱ · ᛊ ᛗ · ᛚ · τ · ᚠ · ᚢ · ᚱ · ᚾ · ᛊ · ᛃ · ᛟ · ᚨ · ᛞ

Begin the invocation

Free. No credit card. No wallet. Just your API key to the decentralized network.

Get Your Free API Key →