OpenAI-compatible · Text · Image · Video · Prepaid credits

OpenAI-compatible API for open models

Run DeepSeek, Qwen, Kimi, SDXL, FLUX, and Wan through one API. Keep your OpenAI SDK code, change one base URL, and pay only for the tokens, pixels, or seconds you use.

Create account View 60-second setup

Prepaid credits from $5 · No monthly subscription · Usage visible after every request

$0.085

per 1M input tokens

DeepSeek V4 Flash

$0.0016

per 1024×1024 image

SDXL text-to-image

$0.15

per 5 second clip

Wan2.2 image-to-video

After signup

The fastest path from visit to first API call

Create an account

Create an API key

Your dashboard shows balance, base URL, keys, and recent usage in one place.

Change the base URL

Keep your OpenAI SDK code and point it at the Crustoff endpoint.

OpenAI SDK compatiblePrepaid spend controlUsage visible after every requestOne key across model families

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_CRUSTOFF_KEY",
    base_url="https://api.crustoff.app/v1",
)

resp = client.chat.completions.create(
    model="deepseek-v4-flash",
    messages=[{"role": "user", "content": "write a launch email"}],
)

image = client.images.generate(
    model="sdxl",
    prompt="clean product photo of a matte black espresso machine",
    size="1024x1024",
)

Why builders try it

Lower the cost of experimentation before you scale traffic.

Crustoff is a practical route for prototypes, internal tools, content pipelines, and product features where open models are good enough and predictable spend matters.

Start testing

One key for every modality

Chat, completions, embeddings, image generation, and short image-to-video calls behind one OpenAI-compatible API.

Transparent unit pricing

Pay per token, per megapixel, or per second with prepaid credits. No subscriptions and no surprise monthly minimums.

Built for quick model swaps

Run open-weight models without vendor lock-in. If the model you need is missing, request it before you commit.

Transparent pricing

Pay for the exact units you use.

Prepaid credits only. Text is billed per token, images per megapixel, and video per second. Image pricing uses the honest per-megapixel rate instead of rounding 1024×1024 up to 2 MP.

Text · chat, completions, embeddings

DeepSeek V4 Flash

fast MoE · 1M ctx · cached in $0.017

$0.085 / $0.17

Qwen3.6 35B-A3B

MoE · 256K ctx

$0.1275 / $0.8075

Qwen2.5 7B Instruct

chat · per 1M in / out tokens

$0.20 / $0.60

GLM-5.2

flagship · 1M ctx · cached in $0.153

$0.8075 / $2.55

Kimi K2.7 Code

coding/agentic · cached in $0.1275

$0.629 / $2.975

DeepSeek V4 Pro

reasoning · 1M ctx · cached in $0.085

$1.105 / $2.21

Image · text-to-image

SDXL

≈ $0.0016 per 1024×1024 image

$0.0015 / MP

FLUX.1 schnell

≈ $0.0019 per image · 1–4 steps

$0.0018 / MP

Qwen-Image

≈ $0.013 per image · flagship quality

$0.012 / MP

Video · image-to-video

Wan2.2

image-to-video · ≈ $0.15 per 5s clip

$0.030 / sec

Create account Request a model

Before you switch

Practical answers for your first request.

Can I use the official OpenAI SDK?

Yes. Keep your existing SDK and request structure, then replace the API key, base URL, and model name. Supported endpoints include chat, embeddings, and image generation.

How does billing work?

Crustoff uses prepaid credits with no monthly subscription. Add credits in your dashboard before making API requests; usage is deducted per token, megapixel, or second.

Which models are available?

The current catalog includes DeepSeek, Qwen, Kimi, SDXL, FLUX, Qwen-Image, and Wan. You can request another open model if it is not listed.

Try the cheaper path before routing more traffic elsewhere.

Sign up, add prepaid credits, create a key, and run a real request. If a model is missing, tell us what you need and we'll prioritize demand.

Create account Read the docs