One value, zero config
Swap a specific model ID for fastrouter/auto and FastRouter's intelligent routing engine takes over model selection on every request.
Automatic Model Selection
Stop hardcoding one model. Let FastRouter choose.
Set your model to fastrouter/auto and FastRouter's routing engine analyzes each request, then picks the best model from the available pool-weighing query complexity, domain, and cost-efficiency, with no manual tuning.
No credit card required · Free to start
Automatic model selection
RequestIntelligent routing
model:"fastrouter/auto"
messages: [{ role: "user", … }]
Routing engine analyzes the request
ComplexityDomainCost
selected_model
anthropic/claude-haiku-4.5
top_candidates
openai/gpt-5.5google/gemini-3-pro
Why automatic model selection
The fastest way to route to the right model
Delegate the model choice to FastRouter and skip researching, pinning, and maintaining a model for every workload.
Chosen on real signals
Each request is matched to a model by the complexity of the query, its domain or topic, and cost-efficiency-not a fixed default.
Nothing to maintain
Skip manual tuning and model preference lists. It's the fastest way to start, and the best model is picked from the available pool.
How it works
From prompt to the best model, automatically
Send your usual chat completion with model set to fastrouter/auto. The routing engine reads the conversation, scores what the request needs, and routes to the best model in the pool-then returns the completion.
Request
Prompt + fastrouter/auto
model: automessages
- Send a standard chat completion with model set to fastrouter/auto.
- Pass your usual messages array-no extra parameters.
FastRouter
Routing engine
ComplexityDomainCost-efficiency
- Analyzes the conversation to score what the request needs.
- Selects the best model from the available pool.
Response
Best-fit model
selected_modeltop_candidates
- Routes the request to the chosen model automatically.
- Returns the completion from the selected model.
One model value
Delegate selection with fastrouter/auto
Instead of picking a specific model, set the model field to fastrouter/auto. FastRouter's intelligent routing mode delegates the decision to its engine while your request stays fully OpenAI-compatible.
Drop-in model value
Send model: fastrouter/auto to /api/v1/chat/completions-no new SDK and no extra parameters to wire up.
Your usual messages
Pass the same messages array; the routing engine analyzes the conversation to determine the optimal model.
No preference lists
There's nothing to curate or tune-skip maintaining a list of models for each workload.
POST /api/v1/chat/completions
JSON{
"model": "fastrouter/auto",
"messages": [
{
"role": "user",
"content": "Explain this stack trace"
}
]
}
Routed to selected_modelBest from the pool
Selection signals
Matched to each request on three signals
Every request is routed on what it actually contains. FastRouter weighs how hard the query is, what it's about, and how to keep it cost-efficient before choosing a model.
Complexity of the query
Heavier reasoning tasks can route to stronger models while simple prompts stay lightweight.
Domain or topic
The subject of the request informs which model is the best fit for the answer.
Cost-efficiency
Selection factors in cost, so you're not overpaying for every call by default.
Selection signals
- Complexity of the queryHigh
- Domain or topicCode
- Cost-efficiencyBalanced
assigned_tags
reasoningcode
Routing engine
Preview the pick before you spend a token
The model selection endpoint returns the recommended model for a prompt without running the completion-so you can inspect exactly what the engine would choose.
Recommended selected_model
The engine returns the single best model ID it would route the request to.
Ranked top_candidates
Inspect the candidate models the engine considered, plus the assigned_tags for the prompt.
Preview without spending
Call /api/v1/model/selection to see the decision without making the actual completion request.
POST /api/v1/model/selection
autotop_candidates considered
- anthropic/claude-haiku-4.5selected_model
- 2openai/gpt-5.5candidate
- 3google/gemini-3-procandidate
Returns the pick without running the completion.
Auto vs hardcoded
fastrouter/auto vs pinning one model
Automatic model selection adapts the model to each request. A hardcoded model ID always runs the same model-handy when you need determinism, but it can't respond to what the request actually needs.
| Behavior | fastrouter/autoAutomatic selection | Hardcoded modelPinned model ID |
|---|---|---|
| Per-request model selection | ||
| Selects the model per request | Included | Not included |
| Adapts to query complexity | Included | Not included |
| Routes by domain or topic | Included | Not included |
| Factors in cost-efficiency | Included | Not included |
| Setup | ||
| Requires picking a model upfront | Not included | Included |
| Best model from the available pool | Included | Not included |
fastrouter/auto delegates the model choice to FastRouter's routing engine; a hardcoded model ID always runs the same model regardless of the request.
Use cases
When to let FastRouter choose
fastrouter/auto fits any workload where the right model depends on the request rather than a fixed, upfront choice.
Ship a prototype fast
Start building without researching models-fastrouter/auto picks a sensible model from day one.
Mixed workloads, one endpoint
Send simple and complex prompts to the same endpoint and let each route to a fitting model.
Spend less without micromanaging
Let cost-efficiency factor into every pick instead of forcing every call through one expensive model.
Drop into OpenAI-compatible apps
Change only the model value to fastrouter/auto-the rest of your chat completions request stays the same.
FAQ
Automatic model selection, answered
Set "model" to "fastrouter/auto" in your chat completions request instead of a specific model ID. FastRouter's intelligent routing mode then delegates model selection to its engine-no other configuration is required.
It selects the most appropriate model based on three signals: the complexity of the query, the domain or topic of the request, and cost-efficiency. The routing engine analyzes your conversation messages to make the call.
FastRouter analyzes the input and selects the best model from the available pool. You don't curate or rank models yourself-the engine picks the best fit for each request.
No. Automatic model selection needs no manual tuning and no model preference lists. It's the fastest way to get started-just send fastrouter/auto and FastRouter handles the choice.
Yes. Call the model selection endpoint at /api/v1/model/selection with model fastrouter/auto and your messages. It returns the recommended selected_model, the top_candidates considered, the assigned_tags for the prompt, and the selection_strategy-without making the actual completion request.
With a specific model ID, every request runs that same model. With fastrouter/auto you delegate the decision per request, so the model adapts to the prompt's complexity, domain, and cost-efficiency. Automatic selection is opt-in through the model value, so you can still pin a model whenever you need a fixed one.
Let FastRouter pick the right model
Swap your model value for fastrouter/auto and route every request to the best model in the pool-no tuning, no preference lists, no extra code.