- Models available
- 93
- All callable with one API key
- Model providers
- 13
- One account, one bill
- Modalities
- 3
- Text · Image · Video
- Tokens a day
- 100B+
- Routed through the gateway
Connected
- OpenAI
- Anthropic
- DeepSeek
- xAI
- QwenCloud
- ByteDance
- Kling
- MiniMax
- Moonshot
- Tencent
- Xiaomi MiMo
- Z.AI
01Unified access
Many models.One integration.
Protocols, credentials, and billing from different providers, behind a single endpoint. Keep the SDKs you know and change only the base URL and API key.
One endpoint
OpenAI, Anthropic, and Gemini protocols share the same base URL, with no per-provider integration.
One key
Authentication, routing, and billing are handled in one place, so you don't open and fund an account per provider.
Switch models freely
Changing models usually means changing the model parameter, so every task gets the right model.
Connection diagram, not live service status
Familiar protocols. More freedom to choose.
OpenAI / Anthropic / Gemini
02Every modality
Text, image, video, and embeddings,on one platform.
From reasoning and chat to multimodal generation, turn the strengths of different models into your next product capability.
Images / API
Give an idea its shape.
From one product image to an entire visual direction, connect image generation and editing to your creative workflow.
Explore models for this capability
Public price for one meter. Full pricing conditions are on the model page.
Original visual study · Not a model benchmark
03Reliability
Upstreams fluctuate.Your service doesn't have to.
The gateway routes every request in real time and switches channels when one fails, so you don't need another layer of retries and failover in your code.
- Select channel2 mscapabilities · breaker · affinity · priority
- Attempt 1 · Channel A118 ms503 upstream overloaded → switching
- Attempt 2 · Channel B312 ms200 · first byte 312 ms
- Streaming responseSSE
Sure. Here are three key points: define the goal, break it into steps, then verify each one.
To your application, this is just a normal streaming response.
Smart routing
Each request goes to a suitable upstream channel based on model capabilities, channel priority, and cost strategy.
Automatic failover
When an upstream times out, rate-limits, or errors, the gateway retries on another available channel before the first byte is returned.
Circuit breaking
Channels that keep failing are isolated temporarily and probed for recovery, so requests stop hitting the same fault.
Cache affinity
Consecutive requests from the same key to the same model stay on the same upstream when possible, so multi-turn chats and agents hit prompt caches more often.
Illustrative only. Channel names and timings are demo data.
04Observable · Controllable
Every callvisible, accountable, and bounded.
Request logs, cost breakdowns, and key limits in one place, ready for production from your first call.
/v1/chat/completions| Request ID | Status | Cost |
|---|---|---|
| 200 | 0.00504 | |
| 200 | 0.002952 | |
| 429 | 0 |
- First token
- 248 ms
- Input tokens
- 1,240
- Output tokens
- 320
One request. The complete picture.
Status, first-token latency, usage and a request ID in one place. Investigate without jumping between provider dashboards.
- Model
- Demo language model
- Status
- Succeeded
- Request total
- 0.00504 credits
Interactive interface illustration. All requests, charges and limits are demo data.
05Quickstart
Three stepsto your first call.
Create an API key, choose a model, and send a streaming request with an SDK you already know.
- 01Create an API key
- 02Choose a model and protocol
- 03Set base_url and run the example
These examples use OpenAI-compatible Chat Completions. See the docs for other protocols and model-specific parameters.
from os import environfrom openai import OpenAI client = OpenAI( base_url="https://api.modelrover.ai/v1", api_key=environ["API_KEY"],) stream = client.chat.completions.create( model="anthropic/claude-fable-5-2", messages=[{"role": "user", "content": "Hello!"}], stream=True,) for chunk in stream: if chunk.choices: print(chunk.choices[0].delta.content or "", end="", flush=True)Install the SDK for your language and set API_KEY in your server environment first.
06FAQ
Before you start,you may want to know.
Key rules for protocol compatibility, billing, and data retention. See the developer docs for details.
The platform supports client protocols including OpenAI, Anthropic and Gemini. For a supported protocol, keep the corresponding SDK and change the endpoint, API key and model ID. Each model's capabilities and parameter support are documented separately.
Read the full guideSpend your time on the product.Leave integration to the gateway.
Create an API key, set base_url, and start calling. Pay for actual usage, with no account to open per provider.
API endpointhttps://api.modelrover.ai/v1