One alias for many models
Group several models and providers under a single alias name, then reference it in the model parameter of any request-just like a normal model id.
Virtual Model Aliases
Swap models without touching your code
Group any set of models and providers under one stable alias, then reference it in the model parameter. FastRouter resolves the alias per request with your chosen strategy and automatic fallback-so you can swap, upgrade, and A/B test models without shipping code.
No credit card required · Free to start
Virtual model alias
Production Chat
model: "prod-chat"
Strategy
Priority routing
Fallback
Automatic
3 models mapped
Resolved in priority order · scoped to Production
- 1openai/gpt-5.5OpenAIPrimary
- 2anthropic/claude-opus-4.8AnthropicFallback
- 3google/gemini-3-proGoogleFallback
Why virtual model aliases
Decouple your app from any single model
Reference one stable name and let FastRouter decide which model, provider, and routing policy actually serves each request.
Routing strategies that fit the job
Choose how each request resolves: priority order, random shuffle, lowest latency, highest throughput, lowest usage, lowest price, or category-based routing.
Reliability built in
If a model or provider fails, FastRouter transparently routes to the next candidate in your list-so traffic keeps flowing without manual intervention.
How it works
One alias, resolved at request time
Your app sends a single model id. FastRouter resolves the alias, applies your selection strategy, and routes to the right model and provider-falling back automatically when one is unavailable.
Your app
One stable model id
model: "prod-chat"
- References the alias name in the standard model parameter.
- No SDK or prompt changes when the targets behind it change.
FastRouter
Alias resolution
Selection strategyWeighted routingAuto-fallback
- Resolves the alias and picks a target with your selection strategy.
- Falls back to the next candidate when a provider is unavailable.
Targets
Models & providers
GPT-5.5Claude Opus 4.8Gemini 3 Pro
- Routes to the chosen model on the chosen provider.
- Add, swap, reweight, or reorder targets from the dashboard.
Alias mapping
One alias, many models and providers
Group the models and providers you trust under a single alias and assign weights where they help. FastRouter picks a target on every request, and any caller referencing the alias gets the new lineup automatically.
Mix models and providers
Map a single alias to several models spanning OpenAI, Anthropic, Google, and other providers.
Weight the traffic
Assign weights so each model takes the share of requests you choose-ideal for gradual rollouts.
Reference by name
Call the alias from the model parameter exactly like any other model id-no new SDK calls.
Alias targets
- openai/gpt-5.550%OpenAI
- anthropic/claude-opus-4.830%Anthropic
- google/gemini-3-pro20%Google
3 models across 3 providers · weighted traffic split
Selection strategies
Choose how every request routes
Each alias carries a selection strategy that decides which target handles a request. Optimize for resilience, cost, speed, load balancing, or task fit-without rewriting your application.
Priority and fallback order
Route in a fixed, user-defined order and fall back deterministically to the next candidate.
Performance and cost aware
Auto-select by lowest latency, highest throughput, lowest usage, lowest price, or weighted random shuffle.
Category-based routing
Send each prompt to a model group by detected category, each with its own sub-strategy and weights.
Selection strategy
7 strategiesPriority routing
Deterministic, ordered fallback
Random shuffle
Weighted A/B distribution
Lowest latency
Fastest current response time
Highest throughput
Highest sustained request rate
Lowest usage
Balance load across targets
Lowest price
Cheapest eligible model
Category-based
Route by detected prompt category
Zero-code swaps
Swap and upgrade models without touching code
Because callers only know the alias, you can change what it points to whenever you want. Promote a new model, add a provider, or reorder fallbacks and deployed apps pick up the change on their next request.
Change targets centrally
Update the alias in the dashboard and every key and project using it follows instantly.
No redeploys required
No SDK upgrade, prompt edit, or release is needed to move traffic to a new model or provider.
A stable contract
Your code keeps sending one model id while the lineup behind it evolves over time.
Your application code
Unchanged{
"model": "prod-chat",alias
"messages": [ … ]
}
prod-chat resolves to
Wasopenai/gpt-5.5
Nowanthropic/claude-opus-4.8
Update the alias in the dashboard-no redeploy and no SDK change, every caller picks it up on the next request.
Alias vs hardcoded model id
Why route through an alias
A virtual model alias is referenced exactly like a model id, but it adds the flexibility, routing control, and reliability a single hardcoded model can't.
| Capability | Hardcoded model idSingle fixed model | Virtual model aliasOne name, many targets |
|---|---|---|
| Change management | ||
| Swap or upgrade models without code changes | Not included | Included |
| Reorder, add, or remove targets centrally | Not included | Included |
| Stable reference across every deploy | Not included | Included |
| Routing & reliability | ||
| Weighted traffic split across models | Not included | Included |
| Strategy-based selection (latency, price, usage) | Not included | Included |
| Automatic fallback to the next candidate | Not included | Included |
| Category-based routing by prompt type | Not included | Included |
An alias is referenced the same way as a model id-by passing its name to the model parameter.
Use cases
Built for teams that change models often
From A/B tests to outage resilience and cost control, aliases let you evolve your model strategy without touching application code.
A/B test models and providers
Split traffic by weight with Random Shuffle to compare models or providers on live requests before committing.
Stay up through provider outages
Lean on built-in fallback so a failed model or provider transparently rolls over to the next candidate in your list.
Optimize for cost or speed
Route to the lowest price, lowest latency, or highest throughput target to match each workload's priorities.
Specialize by task
Use category-based routing to send coding, translation, or classification prompts to the models best suited for them.
FAQ
Virtual model alias questions, answered
A virtual model alias is a custom name that maps to one or more underlying models and providers. You reference the alias by passing its name to the model parameter in your API calls, and FastRouter automatically selects and routes each request to one of the configured targets according to the selection strategy you choose.
Pass the alias name to the model parameter, exactly as you would a normal model id. FastRouter resolves the alias on each request, applies your selection strategy, and routes to the chosen model and provider-so no other code changes are required.
You can route with Random Shuffle (weighted distribution), Lowest Latency, Highest Throughput, Lowest Usage, Lowest Price, Priority Routing (a fixed, ordered fallback), or Category-Based Routing, which sends each request to a model group based on the detected prompt category. Each strategy decides which target handles a given request.
Fallback is built in. If a model or provider fails or is unavailable, FastRouter transparently sends the request to the next candidate in your list. For Priority Routing, fallback follows the numbered order you define; for Category-Based Routing, it follows the sub-strategy configured for that category.
Yes. Because your application only references the alias name, you can add, swap, reorder, or reweight the underlying models and providers from the dashboard at any time. Deployed apps pick up the new configuration on their next request-no redeploy or SDK change needed.
Category-Based Routing classifies each prompt and routes it to a matching model group. You define a Default category that is always present, plus any named override categories such as Coding, Content Writing, or Translation. Each category has its own model list, provider preferences, weights, and sub-strategy, and anything that doesn't match a named category falls through to the Default.
Ship one model id, evolve everything behind it
Create your first virtual model alias, map the models and providers you trust, and pick a routing strategy-in minutes.