HeadsUpAI

OpenRouter Adds Cost Quality Slider to Automate Model Selection Expenses

OpenRouter released a cost_quality_tradeoff parameter for its Auto Router, an intelligent selection tool powered by NotDiamond. The update introduces a 0–10 scale to weigh model capability against price. A value of 0 prioritizes the most capable model regardless of cost, while 10 forces the cheapest available option.
Parameter
cost_quality_tradeoff
Range
0 to 10
Default value
7
Routing engine
NotDiamond
Model pool
GPT-5.1, Claude Sonnet 4.5, Gemini 3.1 Pro, and others

This control addresses the complexity of managing inference (the process of running a trained model to generate outputs) costs. It extends the logic of Auto Exacto and complements the Model Comparison Tool. While Pareto Code targets coding benchmarks, this update provides a general-purpose lever for all prompt types.

Implement the tradeoff by adding the cost_quality_tradeoff integer to the plugins section of an API request or the settings UI. The default is 7, balancing savings with quality. The router selects from a pool including openai/gpt-5.1 and anthropic/claude-sonnet-4.5 at standard rates.

View the full update on openrouter.ai
OpenRouter
OpenRouter
@OpenRouter
X

The Auto Router now lets you tune how it weighs cost against quality. New `cost_quality_tradeoff` parameter, 0 to 10: Set it to 0 and it always picks the most capable model regardless of price. Set it to 10 and the cheapest model wins. https://t.co/nWr4AuwMC2

3retweets45likes
View on X

Still wondering? A few quick answers below.

The Auto Router is an intelligent model selection tool that automatically chooses the best LLM for a specific prompt. It analyzes factors like task type and prompt complexity to route requests to a curated set of high-performance models, including options from OpenAI, Anthropic, and Google.

The parameter uses a 0–10 scale to define routing priorities. A setting of 0 ignores price to select the most capable model available, while a setting of 10 prioritizes the lowest cost. Intermediate values allow users to find an optimal balance for their specific use case.

No, there is no additional fee for using the Auto Router. Users simply pay the standard per-token rate for whichever model the system selects to handle the request. This allows for dynamic cost optimization without adding overhead to the existing OpenRouter pricing structure.

Yes, users can configure allowed models using wildcard patterns in the API request or the settings UI. For example, you can limit the router to only select from Anthropic models or specific versions of GPT, ensuring the system stays within preferred provider or capability boundaries.

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

Share this update