Liner, an AI agent startup, released a new solution that automatically routes user queries to appropriate large language models, targeting the cost inefficiencies that plague enterprise AI deployments.
The Challenge of AI Cost Efficiency
The company unveiled the Liner Model API, which analyzes query difficulty and type before assigning an LLM suited to the task. According to Liner, the approach can reduce token costs by more than 50 percent while maintaining answer quality comparable to responses from high-performance models.
The core problem Liner aims to address: current enterprise AI systems send all queries through top-tier models regardless of complexity. Simple factual questions and basic reasoning tasks consume the same resources as advanced coding or mathematical problems, creating what the company describes as unnecessary infrastructure expenses.
Query Routing Distribution
Under the Liner Model API framework, approximately 43 percent of queries in internal testing went to high-performance models. These handled tasks requiring coding, math, and complex reasoning. The remaining 57 percent were processed through more cost-efficient models optimized for general queries.
The system requires no additional implementation overhead. Companies can integrate it by swapping in the new API to replace existing infrastructure, without building custom routing logic.
Technology Development
Liner developed the orchestration technology using its proprietary model evaluation system, drawing on data accumulated through actual service operations. Performance verification relied not only on public benchmarks but also on real user query data.
“Handling every question only with top models is like bringing in a supercomputer for every simple calculation,” said Jin-woo Kim, CEO of Liner. He noted that the tool eliminates the ongoing burden for developers to evaluate and replace models as new options enter the market with varying performance and pricing profiles.
The company positions the Liner Model API as a practical solution for organizations looking to optimize their AI spending without sacrificing response quality on complex tasks.