WHO ARE YOU?

Use AI with a speed guarantee. If the response is late, you get paid back automatically — no complaints, no forms.

AI that pays you back
if it's too slow

Pick a speed tier. If the AI responds too slowly, money is automatically sent from the worker's wallet to yours — no waiting, no dispute, no customer support.

SLA PASS RATE
0.0%
P99 TTFT
0ms
PENALTIES PAID
0 $LOCK
ACTIVE WORKERS
0
132 TEE-capable
IN PLAIN ENGLISH

What is Gridlock, really?

Think of it like Uber for AI compute but the car has a guaranteed arrival time, and if it's late, you automatically get a refund.

Anyone can be a worker
Have a gaming PC with a strong GPU? You can earn money by running AI on it. You install software, register on Gridlock, and your computer starts processing AI requests from people around the world.
Speed is guaranteed by code
When you request AI, you pick a speed tier. The worker's $LOCK tokens are locked as collateral. If the AI is too slow, the blockchain automatically moves tokens from the worker's wallet to yours. No humans involved.
No one controls it
Everything runs on Solana — a public blockchain. There is no company in the middle that can raise prices, cut you off, or disappear. The rules are written in code that anyone can read.
HOW A REQUEST FLOWS
You send a prompt
via API or Console
Router picks a worker
best GPU for your tier
Worker runs the AI
on their GPU
Response + proof
delivered to you
Penalty or payment
settled automatically
FOR DEVELOPERS

One URL change. Guaranteed latency.

Gridlock is a drop-in replacement for the OpenAI API. If the request misses your SLA target, the penalty is auto-credited on-chain — no dispute, no claim, no waiting.

1
Point your SDK at api.grid-lock.tech
2
Add sla: 'realtime' to your request
3
Get paid back if it's late
// Change one URL, get a latency guarantee
import OpenAI from 'openai'
const client = new OpenAI({
baseURL: 'https://api.grid-lock.tech/v1',
apiKey: process.env.GRIDLOCK_KEY,
})
const res = await client.chat.completions.create({
model: 'llama-3.1-70b',
messages: [{ role: 'user', content: '...' }],
gridlock: { sla: 'realtime' },
})
// { ttftMs: 187, slaMet: true }
SPEED TIERS

Pick how fast you need it

Miss the target — penalty auto-pays from worker stake directly to your wallet.

REALTIME
300ms
first response target
Penalty if late: 2× fee
Chatbots, voice agents
STANDARD
800ms
first response target
Penalty if late: 1× fee
RAG apps, copilots
BATCH
5s
first response target
Penalty if late: 0.25× fee
Summarization, embeddings
CONFIDENTIAL
800ms
first response target
Penalty if late: 1× + slash
Enterprise / private prompts
COMPETITIVE POSITION

The only one that sells a guarantee

FeatureRenderio.netAkashStratumGRIDLOCK
1.Enforceable latency SLA————YES
2.Penalty auto-payout————YES
3.Disaggregated Prefill/Decode————YES
4.KV-cache routing———PARTYES
5.Confidential TEE serving————YES
6.Goodput optimization———PARTYES

Ready to build?

Change one URL. Add one field. Get a latency guarantee backed by real money.

Open SLA ConsoleRead the Docs

Penalties are automatic. No trust required.