The open exchange for AI inference

AI inference.
Use it now. Trade it forward.

Call models through one API. Pay as you go, or buy monthly OSTR to cover the usage you can plan for.

Zero Data RetentionNo prompts or outputs stored by default.
Use API Trade OSTR

One API key. Pay-as-you-go access. Monthly forward series.

AI Gateway

ModelInput / MOutput / MLatencyUptime
GLM-5.3$1.40$4.40~1158ms100%
DeepSeek V4 Pro$0.66$1.98~2568ms100%
Qwen3.8 Flash$0.15$0.47~45ms100%
GLM-5.2$1.40$4.40~783ms100%
Kimi K3$3.00$15.00~332ms100%

AI Gateway

Your agents.
One gateway.

Bring leading models into the tools you already use. One API key, one balance, pay as you go.

Explore all models Explore AI Gateway

Forward Market · OSTR

Trade inference.
Power your agents.

Each OSTR provides $20 of inference credit for a defined month. Use it across eligible models or sell unused OSTR before the serving period ends.

Trade OSTR

Live Series

$20 inference credit per OSTRMonthly series · Inference delivery

Meet OSTR

One balance.
Multiple models.

OSTR is a tradable right to AI inference for a defined month. Each series sets the eligible models and fixed prices. Each OSTR provides $20 of inference credit.

Explore OSTR

01 / THE MONTH

Choose your serving month.

Each series has a defined start and end date. Use your inference within that window.

02 / THE MENU

Choose the model for each call.

Call eligible models through the Mercatus API. Your series’ menu lists their fixed prices.

03 / THE CREDIT

Know what your OSTR buys.

Each OSTR provides $20 of inference credit at your series’ fixed model prices. Buy at $16 and save 20% when you use the full credit.

04 / THE TRADING

Trade your remaining balance.

Sell unused OSTR before the serving period ends. A matching buyer is required.

Unused OSTR expires at the end of the serving window. No automatic rollover or cash redemption.

See series terms

Why Mercatus

More choice.
More control.

Access models, manage inference purchases, and follow market prices in one place.

MODEL ACCESS

More models. One integration.

Reach models across providers with one API key and one account balance. Choose the model that fits your workload.

Read the quickstart

FLEXIBLE PROCUREMENT

Trade as your needs change.

Use pay-as-you-go access for variable demand. Buy OSTR for planned usage, with the option to offer unused units for resale before expiry.

Explore OSTR

PUBLISHED FEES

Clear prices. Clear fees.

No Mercatus markup on spot prices. Forward trading fees are published separately, with no commission on inference delivery.

See fees

MARKET BENCHMARKS

Put prices in context.

Follow token and GPU price benchmarks alongside market quotes to inform your inference purchases.

Explore the indices

Daily benchmarks

Follow the cost of AI.

Track model inference and GPU rental prices with daily benchmarks. Explore trends across model categories and GPU types.

These indices are reference benchmarks, not tradable contracts.

TPI · Token Price Index

137 models

AI inference prices

Blended USD per 1M tokens, by model category.

US TPI

$5.07+27.88%

30-day trend

  • US TPI$5.07+27.88%
  • CN TPI$1.24+8.06%
  • Coding TPI$2.14+5.42%
View TPI

GPI · GPU Price Index

40 providers

GPU rental prices

On-demand USD per GPU-hour, by GPU model.

H100

$4.00/hr+4.53%

30-day trend

  • H100$4.00/hr+4.53%
  • H200$5.01/hr+12.81%
  • B200$6.81/hr+6.76%
  • B300$8.93/hr+0.40%
View GPI

Get started

Your first call,
in three steps.

Connect your app or agent to Mercatus and start with pay-as-you-go inference.

Python · Responses API

from openai import OpenAI

client = OpenAI(
    base_url="https://gateway.mercatus-ai.com/v1",
    api_key="YOUR_MERCATUS_KEY",
)

with client.responses.create(
    model="claude-opus-4-7",
    input="Hello!",
    stream=True,
) as stream:
    for event in stream:
        if event.type == "response.output_text.delta":
            print(event.delta, end="", flush=True)

  1. Create your API key.

    Sign in and create an account-level key. Keep it on your server and use it across the models you call.

    Open API Keys

  2. Fund your balance.

    Add funds to pay for API usage. Start with the models and workloads you need today.

    Open your dashboard

  3. Send your first request.

    Start with the Responses API. Add your key and select a text model from the AI Gateway. This example streams text as it arrives. Allow time for the first chunk.

    Follow the quickstart
Use API Read the docs

For procurement support, talk to sales