anypick
An open-source LLM model selection library available in both Python and TypeScript implementations. It supports automatic synchronization of model catalogs from platforms such as OpenRouter and uses a multi-dimensional filtering pipeline—covering price, latency, benchmarks, and capabilities—to automatically identify the most suitable LLM.
Timeline (last 90 days)
anypick is an open-source library that provides both Python and TypeScript implementations, with the API design kept consistent across both languages.
anypick supports automatically pulling model catalogs from two data sources: OpenRouter and Vercel AI SDK.
anypick adopts a pipeline mode that can chain multiple filter conditions together to progressively narrow down the range of candidate models.
anypick supports filtering models by price, latency, benchmark scores (such as MMLU and HumanEval), and capability dimensions (such as function calling, visual input, and long context).
anypick can select the optimal model from a candidate list based on specified criteria—for example, best_price selects the cheapest model and best_throughput selects the model with the highest throughput—and also supports custom scoring functions.
The design goal of anypick is to eliminate the need to manually swap the underlying model in LLM systems every three months.
Decoupling model selection logic from business logic means that when a model raises its price or is deprecated, simply adjusting the filter conditions will automatically adapt to the next best option.
anypick currently only covers OpenRouter and Vercel, and does not include enterprise-common platforms such as AWS Bedrock and Azure OpenAI.
Dual implementations in Python and TypeScript mean that every feature update requires maintaining two codebases in sync, which is a burden on maintainers.
The timeliness of anypick's data depends on the update frequency of the model catalog; if real-time synchronization is not possible, filter results may be based on outdated data.
All Facts (10)
anypick is an open-source library that provides both Python and TypeScript implementations, with the API design kept consistent across both languages.
50%Unverifiedanypick supports automatically pulling model catalogs from two data sources: OpenRouter and Vercel AI SDK.
50%Unverifiedanypick adopts a pipeline mode that can chain multiple filter conditions together to progressively narrow down the range of candidate models.
50%Unverifiedanypick supports filtering models by price, latency, benchmark scores (such as MMLU and HumanEval), and capability dimensions (such as function calling, visual input, and long context).
50%Unverifiedanypick can select the optimal model from a candidate list based on specified criteria—for example, best_price selects the cheapest model and best_throughput selects the model with the highest throughput—and also supports custom scoring functions.
50%UnverifiedThe design goal of anypick is to eliminate the need to manually swap the underlying model in LLM systems every three months.
50%UnverifiedDecoupling model selection logic from business logic means that when a model raises its price or is deprecated, simply adjusting the filter conditions will automatically adapt to the next best option.
50%Unverifiedanypick currently only covers OpenRouter and Vercel, and does not include enterprise-common platforms such as AWS Bedrock and Azure OpenAI.
50%UnverifiedDual implementations in Python and TypeScript mean that every feature update requires maintaining two codebases in sync, which is a burden on maintainers.
50%UnverifiedThe timeliness of anypick's data depends on the update frequency of the model catalog; if real-time synchronization is not possible, filter results may be based on outdated data.
50%