Global AI Token Comparison Hub
CN
Back to all plans

OpenRouter API

340+ models · OpenAI compatible

OpenAI-compatible unified API for 340+ models at direct provider rates, plus six OpenRouter routers for auto-selection, free routing, and multi-model deliberation

Token API
SubscriptionToken API
Official

Core models

Auto RouterFree RouterFusion RouterPareto Code RouterBody BuilderOwl Alpha
Auto Router

Model id: openrouter/auto. NotDiamond picks the optimal underlying model per prompt—no extra router fee; response includes actual model field, with session_id sticky routing and cost_quality_tradeoff (0–10).

Free Router

Model id: openrouter/free. Smart random selection from free pool—$0 tokens, subject to daily limits.

Fusion Router

Model id: openrouter/fusion. Parallel panel + judge synthesis—cost equals sum of participating model fees.

Pareto Code Router

Model id: openrouter/pareto-code. Coding model shortlist ranked by Artificial Analysis coding scores.

Body Builder

Model id: openrouter/bodybuilder (beta). Natural language to structured API request objects.

See the official site for more models

Additional core model names still appear above, with full details on the latest official page.

Open official site

Plan details

Pay-as-you-go

Pay-as-you-goRecommended
Input
$0.15
Output
$0.60
Official
Usage
Call explicit model ids (e.g. openai/gpt-4o-mini, anthropic/claude-sonnet-4) at provider list prices deducted from Credits. Budget models from $0.01/M input; flagships like Claude Opus up to $15/$75 per 1M tokens.
Models
Provider ordering and fallbacks switch automatically on provider errors—no code changes. Use :nitro for latency, :floor for cheapest provider.
Highlights
One API key covers all listed models from OpenAI, Anthropic, Google, Meta, DeepSeek, Qwen, xAI, and more—full catalog at openrouter.ai/models.
Best for
Production apps needing explicit model ids and precise per-token cost control

Free Models

Free models
Input
$0
Output
$0
Official
Usage
Use openrouter/free router or append :free to any model id—$0 token cost but daily caps apply (50/day default, 1000/day after $10+ credits).
Models
openrouter/owl-alpha is also $0 per token with native tool use and long context—suited to agent workflow testing.
Best for
Prototypes, demos, and zero-cost exploration

Smart Routers

Smart routers
Input
By routed model
Output
auto · fusion · pareto-code · bodybuilder
Official
Usage
openrouter/auto: meta-model picks the optimal underlying model when you're unsure—metadata returns actual model id, billed at selected model's token rate.
Models
openrouter/fusion: parallel expert panel + judge synthesis in one call—cost equals sum of participating model token usage, for high-stakes multi-angle verification.
Highlights
openrouter/pareto-code: auto-selects strong coding models by benchmark score; openrouter/bodybuilder (beta): natural language to structured API request objects.
Best for
Auto-selection, multi-model deliberation, and coding routing scenarios

Notes

  • Third-party models (GPT, Claude, Gemini, etc.) use provider/model ids at models page rates—see each vendor's direct API page on this site.
  • Router responses include the actual underlying model id in metadata; Auto/Fusion/Pareto/Body Builder bill at the routed model's token rates (summed for multi-model calls).
  • openrouter/auto is powered by NotDiamond with no extra router fee—billed at the selected underlying model's token rate; supports session_id sticky routing and allowed_models allowlists.
  • Free model API defaults to 50/day, 1000/day after $10+ Credits; paid model rate limits depend on balance—see docs/faq.

Supported coding tools

REST APIOpenAI SDKStreaming SSETool CallingServer ToolsStructured Outputs

Pricing and model data sourced from official vendor websites

FAQ

General·6
General
6 条