Use case
Use this comparison when your team already knows the image or video workload it wants to run and needs a simpler production gateway around standard model names, endpoint examples, and spend records.
Standard model names and pricing
Source: GET https://cn.crazyrouter.com/api/pricing, snapshot 2026-06-06.
When Crazyrouter is the better fit
Choose Crazyrouter when you want image, video, chat, and embedding calls behind one operational API layer rather than a model-marketplace workflow. API requests use https://cn.crazyrouter.com, while login, recharge, API keys, and usage records stay on https://crazyrouter.com.
- Standard model names are copied directly from the Crazyrouter pricing catalog.
- OpenAI-compatible image calls use POST https://cn.crazyrouter.com/v1/images/generations.
- Video creation uses documented Crazyrouter video paths such as POST https://cn.crazyrouter.com/v1/video/create.
Where fal.ai remains strong
fal.ai is still a strong choice when you need fal-specific model endpoints, marketplace-native examples, or custom serverless model deployment. Crazyrouter should not be treated as a replacement for every fal Serverless use case.
Migration path
Start with one workload, usually image generation or short video generation. Pick a standard Crazyrouter model name, copy the endpoint, run a small request set, then compare accepted-output cost in the Crazyrouter console before moving recurring traffic.
- gpt-image-2: $0.0377/image after 0.65x discount; public_endpoint_types = image-generation.
- nano-banana-2: $0.0369/image after 0.55x discount; public_endpoint_types = image-generation.
- qwen-image-max: $0.0714/image; public_endpoint_types = image-generation.
- veo-3.1-fast: $0.10/second; public_endpoint_types = unified-video.
Decision summary
Use fal.ai for marketplace exploration and custom serverless execution. Use Crazyrouter when production needs clear model rows, consistent API authentication, visible billing, and a gateway your backend can keep stable.
cn.crazyrouter.com test evidence
Only checks that returned 200 are shown here. API requests use https://cn.crazyrouter.com; account, billing, and console actions use https://crazyrouter.com.
| Check | Request | Status | Result |
|---|---|---|---|
| Pricing API reachable | GET https://cn.crazyrouter.com/api/pricing | 200 | Returned the public model pricing catalog used by these pages. |
| Target models visible to API key | GET https://cn.crazyrouter.com/v1/models | 200 | Authenticated check confirmed gpt-image-2, veo-3.1-fast, veo-3.1, nano-banana-2, nano-banana-pro, nano-banana, grok-4-image, qwen-image-plus, qwen-image-max, and qwen-image-2.0 are visible to the local Crazyrouter API key. |
Implementation examples
curl -X POST https://cn.crazyrouter.com/v1/images/generations \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "gpt-image-2",
"prompt": "A clean product photo of a black wireless headset on a white background",
"n": 1
}'from openai import OpenAI
client = OpenAI(
api_key="YOUR_API_KEY",
base_url="https://cn.crazyrouter.com/v1",
)
response = client.images.generate(
model="gpt-image-2",
prompt="A clean product photo of a black wireless headset on a white background",
n=1,
)
print(response.data[0].url)