Run AI
with an API.

Run leading image and video models.
All with one line of code.

const response = await fetch("https://api.mlmux.com/v1/images/generations", {
  method: "POST",
  headers: {
    Authorization: "Bearer " + process.env.MLMUX_API_KEY,
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    model: "gpt-image-2",
    prompt: "a soft armchair shaped like a peeled banana",
    route_mode: "official-cheap"
  })
});

console.log(await response.json());

AI-generated soft armchair shaped like a peeled banana

a soft armchair shaped like a peeled banana.

OpenAI / GPT Image 2

With MLMux you can

OpenAI

GPT Image 2

High-fidelity images with reliable text

Run model

Google

Nano Banana 2

Fast generation with flexible references

Run model

Google

Nano Banana Pro

Professional composition and 4K output

Run model

One API. Seven focused models.

Build with the models
your product needs.

OpenAIGPT Image 2Image GoogleNano Banana 2Image GoogleNano Banana ProImage ByteDanceSeedance 2.5Video ByteDanceSeedance 2.0Video ByteDanceSeedance 2.0 FastVideo ByteDanceSeedance 2.0 MiniVideo

How it works

Start with any model through one consistent API. As your product grows, switch models, automate results, and trace every request without changing your integration.

Run models

MLMux gives you pre-release image and video models behind one endpoint. Choose a model, send its inputs, and receive a consistent task response.

Explore models

import json
import os
import urllib.request

request = urllib.request.Request(
    "https://api.mlmux.com/v1/images/generations",
    data=json.dumps({
        "model": "gpt-image-2",
        "prompt": "a quiet city after rain"
    }).encode(),
    headers={"Authorization": "Bearer " + os.environ["MLMUX_API_KEY"]}
)

print(urllib.request.urlopen(request).read().decode())

Switch models with one API

Use the same authentication, task lifecycle, and error format everywhere. Move between image and video models by changing the model name and inputs.

Read the API docs

const response = await fetch("https://api.mlmux.com/v1/videos/generations", {
  method: "POST",
  headers: {
    Authorization: "Bearer " + process.env.MLMUX_API_KEY,
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    model: "seedance-2.5",
    prompt: "a product film in soft morning light",
    duration: 5,
    ratio: "16:9",
    async: true
  })
});

Track every request

Follow generation with a task ID, inspect latency and cost in usage logs, or receive the completed asset automatically through a webhook.

Open the console

Tasktask_8J4C2P

Modelseedance-2.5

StatusSUCCEEDED

Duration24.9 s

Scale on MLMux

Move from a first generation to production traffic with one stable API. Your team can ship image and video features without operating a different integration for every model.

Compare model pricing

ByteDanceOpenAIGoogleSeedanceGPT ImageNano BananaOne API

Example capacityIllustrative

09:0024%12:0082%15:0046%18:0068%

Automatic scale

Send requests when demand arrives. MLMux keeps the public task API consistent as volume changes, so your application code stays the same.

Seedance 2.5 · 1080p$0.092 / sec

GPT Image 2 · 2K$0.050 / image

Nano Banana 2 · 2K$0.060 / image

View all pricing

Pay for what you use

Every released model has a visible preview rate. Completed tasks appear in usage logs with the model, latency, status, and cost attached.

POST/v1/tasks200 OK

{
  "model": "seedance-2.5",
  "status": "queued",
  "task_id": "task_8J4C2P"
}

Forget model-specific infrastructure

Use one authentication method, one task lifecycle, and one response format across the image and video models in the MLMux catalog.

Example traffic54 RPM

task_8J4C2P24.9 sSUCCEEDED

task_8J4B9N8.4 sSUCCEEDED

task_8J49LT—RUNNING

Usage logs & monitoring

Inspect live RPM, latency, model usage, cost, and failures from the console, then trace a specific task whenever something needs attention.

Imagine

A cinematic poolside scene generated with an image model

what

you

A warm editorial bookshop scene generated with an image model

can

A banana-shaped armchair generated in the MLMux playground — Image playground

A friendly studio robot generated with an image model

build.

{
  "task_id": "task_8J4C2P",
  "model": "gpt-image-2",
  "status": "succeeded",
  "output": ["https://cdn.mlmux.com/result.png"]
}

Build with MLMux and turn a product idea into an image or video workflow through one clean API.

Get started

Run AI
with an API.

Run leading image and video models.
All with one line of code.

Get started for free

const response = await fetch("https://api.mlmux.com/v1/images/generations", {
  method: "POST",
  headers: {
    Authorization: "Bearer " + process.env.MLMUX_API_KEY,
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    model: "gpt-image-2",
    prompt: "a soft armchair shaped like a peeled banana",
    route_mode: "official-cheap"
  })
});

console.log(await response.json());

a soft armchair shaped like a peeled banana.

OpenAI / GPT Image 2

With MLMux you can

OpenAI

GPT Image 2

High-fidelity images with reliable text

Run model

Google

Nano Banana 2

Fast generation with flexible references

Run model

Google

Nano Banana Pro

Professional composition and 4K output

Run model

One API. Seven focused models.

Build with the models
your product needs.

OpenAIGPT Image 2Image GoogleNano Banana 2Image GoogleNano Banana ProImage ByteDanceSeedance 2.5Video ByteDanceSeedance 2.0Video ByteDanceSeedance 2.0 FastVideo ByteDanceSeedance 2.0 MiniVideo

How it works

Start with any model through one consistent API. As your product grows, switch models, automate results, and trace every request without changing your integration.

Run models

MLMux gives you pre-release image and video models behind one endpoint. Choose a model, send its inputs, and receive a consistent task response.

Explore models

import json
import os
import urllib.request

request = urllib.request.Request(
    "https://api.mlmux.com/v1/images/generations",
    data=json.dumps({
        "model": "gpt-image-2",
        "prompt": "a quiet city after rain"
    }).encode(),
    headers={"Authorization": "Bearer " + os.environ["MLMUX_API_KEY"]}
)

print(urllib.request.urlopen(request).read().decode())

Switch models with one API

Use the same authentication, task lifecycle, and error format everywhere. Move between image and video models by changing the model name and inputs.

Read the API docs

const response = await fetch("https://api.mlmux.com/v1/videos/generations", {
  method: "POST",
  headers: {
    Authorization: "Bearer " + process.env.MLMUX_API_KEY,
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    model: "seedance-2.5",
    prompt: "a product film in soft morning light",
    duration: 5,
    ratio: "16:9",
    async: true
  })
});

Track every request

Follow generation with a task ID, inspect latency and cost in usage logs, or receive the completed asset automatically through a webhook.

Open the console

Tasktask_8J4C2P

Modelseedance-2.5

StatusSUCCEEDED

Duration24.9 s

Scale on MLMux

Move from a first generation to production traffic with one stable API. Your team can ship image and video features without operating a different integration for every model.

Compare model pricing

ByteDanceOpenAIGoogleSeedanceGPT ImageNano BananaOne API

Example capacityIllustrative

09:0024%12:0082%15:0046%18:0068%

Automatic scale

Send requests when demand arrives. MLMux keeps the public task API consistent as volume changes, so your application code stays the same.

Seedance 2.5 · 1080p$0.092 / sec

GPT Image 2 · 2K$0.050 / image

Nano Banana 2 · 2K$0.060 / image

View all pricing

Pay for what you use

Every released model has a visible preview rate. Completed tasks appear in usage logs with the model, latency, status, and cost attached.

POST/v1/tasks200 OK

{
  "model": "seedance-2.5",
  "status": "queued",
  "task_id": "task_8J4C2P"
}

Forget model-specific infrastructure

Use one authentication method, one task lifecycle, and one response format across the image and video models in the MLMux catalog.

Example traffic54 RPM

task_8J4C2P24.9 sSUCCEEDED

task_8J4B9N8.4 sSUCCEEDED

task_8J49LT—RUNNING

Usage logs & monitoring

Inspect live RPM, latency, model usage, cost, and failures from the console, then trace a specific task whenever something needs attention.

Imagine

what

you

can

build.

{
  "task_id": "task_8J4C2P",
  "model": "gpt-image-2",
  "status": "succeeded",
  "output": ["https://cdn.mlmux.com/result.png"]
}

Build with MLMux and turn a product idea into an image or video workflow through one clean API.

Get started

Run AIwith an API.

GPT Image 2

Nano Banana 2

Nano Banana Pro

Build with the modelsyour product needs.

Run models

Switch models with one API

Track every request

Automatic scale

Pay for what you use

Forget model-specific infrastructure

Usage logs & monitoring

Run AIwith an API.

GPT Image 2

Nano Banana 2

Nano Banana Pro

Build with the modelsyour product needs.

Run models

Switch models with one API

Track every request

Automatic scale

Pay for what you use

Forget model-specific infrastructure

Usage logs & monitoring

Run AI
with an API.

Build with the models
your product needs.

Run AI
with an API.

Build with the models
your product needs.