The inference market built for surplus
One key for every model. Credits from $SURPLUS stakers, up to 20% off, paid in USDG.
Market › Buy
Market / Buy
Your inference, at a discount.
Live order book on Robinhood Chain. Prices update every 30 seconds.
- Credits in the book
- $0
- Face value, ready to fill
- Deepest discount
- 0%
- Fills first, FIFO per level
- Models on one key
- 464
- OpenAI-compatible API
- Added latency
- <50 ms
- Thin relay, no rewrites
Order book depth
deepest discount fills first- The book is empty right now.
Buy credits
up to 20% offPaid in USDG on Robinhood Chain. No subscription.
- Robinhood Chain
- Paxos
- USDG
- Anthropic
- OpenAI
- xAI
- DeepSeek
- x402
- Blockscout
One market for buyers, sellers, stakers and builders.
Explore the protocol- Buyers
- Sellers
- Stakers
- Builders
- USDG
- CREDIT
- Gateway
- Order book
A credit has a clear state
- Minted01
- Listed02
- Filled03
- Activated04
- Spent05
The same models, for less.What changes is the price, never the answer.
Straight relay
Your request goes to the model you named. Nothing is swapped or downgraded.
Provider's own answer
Responses carry the provider's model name and id, so you can check.
Zero data retention (ZDR)
Inference routes only to ZDR endpoints. We keep billing metadata, never prompts or responses.
A real order book, settled in USDG.Every discount comes from stakers selling spare CREDIT.
Liquidity Book
$0 availableRecent buys
last 0- No buys yet.
Buy, key, request.Three steps from a USDG balance to a cheaper inference bill.
- 01
01 / Step 1
Buy credits in USDG
Pick an amount. The order book fills from the deepest discount first, settled on Robinhood Chain in under a second.
- 02
02 / Step 2
Get one API key
Your credits become API balance behind a single key that reaches every model on the catalogue.
- 03
03 / Step 3
Send requests
Point your OpenAI or Anthropic client at Surplus. Spend is metered per request at each model's discounted rate.
Same models. Less spend.The prompt you already send, billed below list price.
Your prompt
Surplus API keysk-surplus-••••7x9k
The answerReady
Your answer will appear here.
List price
$50.00
With Surplus
$40.00
An open market for useful inference.Pick the work. One key covers every model it needs.
Every model, one key.464 models, below what the lab charges.
Claude Fable 5.1
anthropic/claude-fable-5.1
$50.0$40.0−20%GPT-6 Astra
openai/gpt-6-astra
$50.0$40.0−20%Claude Opus 5
anthropic/claude-opus-5
$25.0$20.0−20%Gemini 3.8 Flash
google/gemini-3.8-flash
$3.75$3.00−20%Grok 4.6
x-ai/grok-4.6
$6.00$4.80−20%Claude Sonnet 5
anthropic/claude-sonnet-5
$10.0$8.00−20%DeepSeek V4 Pro
deepseek/deepseek-v4-pro-0813
$1.98$1.58−20%Kimi K3
moonshotai/kimi-k3
$15.0$12.0−20%
Performance has a measurable floor.A thin relay, dollar settlement and a discount from real supply.
<50ms
Added latency per request
20%
Deepest discount in the book
$0
Setup or subscription fees
<1s
Settlement finality on Robinhood Chain
Built around honest inference.Clear rules for every side of the market.
- 01
01 / Buyer
Pay less for the same model
Buy CREDIT below face value in USDG. One key reaches every model at the provider's own rate.
- 02
02 / Seller
Sell what you don't use
List unused CREDIT at a discount level. Buyers fill the deepest level first, FIFO within it.
- 03
03 / Staker
Earn inference by holding
Stake $SURPLUS and earn CREDIT every hour, pro rata by stake and time.
- 04
04 / Gateway
Relay, never rewrite
Requests go to the model you named. Responses carry the provider's own model id.
- 05
05 / Privacy
Keep prompts private
Zero data retention endpoints only. Billing metadata is kept, prompts and responses are not.
- 06
06 / Settlement
Settle in the open
Every fill, claim and activation is a transaction on Robinhood Chain you can verify.
Analytics you already know.Spend by day, by model, and the latency Surplus adds.
Spent
$39.74
Requests
1.5k
Tokens
3.5M
Latency
31ms
p95 47 ms
Models
- Claude Fable 5.1$18.28696 req
- GPT-6 Astra$10.73409 req
- Gemini 3.8 Flash$5.56212 req
- Grok 4.6$3.18121 req
- DeepSeek V4 Flash 0731$1.9976 req
The whole migration is two lines.Change the base URL and the key. Nothing else.
# before — OpenRouterOPENAI_BASE_URL=https://openrouter.ai/api/v1OPENAI_API_KEY=sk-or-v1-…# after — Surplus. Same two variables: your OpenAI-compatible client reads these.OPENAI_BASE_URL=https://api.surplusllm.com/api/v1OPENAI_API_KEY=sk-surplus-…
Questions, answered.Everything about buying and using Surplus credits.
What am I buying?
A fixed amount of API credit. Enter the value of credits you want, then review the discounted price and platform fee. For example, $100 in credits gives you $100 of model usage through your Surplus key.
How is the platform fee calculated?
The platform fee is 5% of the discounted credit price. The purchase card shows the credit price, fee and total separately, along with your discount after the fee.
Where does the discount come from?
From SURPLUS stakers who offer unused CREDIT at a discount on Robinhood Chain.
Am I getting the real model?
Yes. The gateway is a relay: your request goes to the model you named, unchanged, and the response comes back with the provider's own model name and id on it. Nothing is substituted or downgraded.
Do you store my prompts?
No. The gateway keeps billing metadata only: the model, the token counts, the cost and the timing of each request. Prompts and responses pass through and are not written anywhere.
Whose key serves my requests?
Requests use a managed upstream provider account through the Surplus Gateway. You use your own Surplus API key.
What do I change in my code?
The base URL and the key. Model names, streaming, tool calls and vision work as they do today.
Is there a subscription or a minimum?
No subscription. Choose at least $5 in credits. Purchased credits become API balance, which you spend as you use your key.
What does Surplus add to each request?
Your account page shows the measured gateway overhead for your requests.
Can I get my money back?
Credit is a grant of product access and is not redeemable for cash. If a payment goes wrong, reply to the receipt and we will sort it out.