arxiv arXiv cs.AI · 1d ago · src: 4d ago · research

AgentRouter reduces agentic workflow costs by 72% via step-level model routing

from English

Researchers propose AgentRouter, a lightweight classifier that optimizes multi-step agentic workflows by routing individual trajectory steps to appropriate model tiers rather than using frontier models for every task. The system addresses the inefficiency of existing solutions that ignore varying subtask complexity within a single agent session.

AgentRouter uses a 12M parameter model with less than 5ms overhead per step on an A100 GPU to map steps to four tiers based on five features.
Trained on 50,000 annotated steps across planning, coding, research, and data analysis tasks, it achieves 72% cost reduction while retaining 97.3% of frontier-only quality.
Routing accuracy reaches 91% on minimal-complexity steps and 85% on efficient-tier steps, outperforming RouteLLM and FrugalGPT which achieved only 31% and 44% cost reductions respectively.

This approach significantly lowers inference costs for enterprise agentic systems by recognizing that different steps within a trajectory require different levels of model capability.

Importance 2/3 New harness with differentiators arXiv cs.AI AI agents Inference efficiency

Related articles

arxiv arXiv cs.AI · 4h ago · 9 views

CliffCompaction reduces coding agent costs by 50% while maintaining performance

Researchers introduce CliffCompaction, an autocompaction technique designed to reduce costs for long-horizon coding agents by up to 50% under a bounded context. The method maintains or improves performance on Terminal-Bench and achieves state-of-the-art results on KernelBench by keeping compacted information faithful through truncation rather than rephrasing.

arxiv arXiv cs.LG · 5h ago · 9 views

CliffCompaction reduces long-horizon coding agent costs by up to 50% while maintaining performance

Researchers introduce CliffCompaction, an autocompaction technique designed for agents handling millions of tokens, which cuts costs by up to 50% under bounded context. The method maintains or improves performance on Terminal-Bench and achieves state-of-the-art results on KernelBench by keeping information faithful through truncation rather than rephrasing.

lab OpenAI News · 11h ago · 16 views

Parallel halves research time and cost using GPT-6 Astra

Parallel has integrated OpenAI's GPT-6 Astra into its AI agent infrastructure, achieving a 50% reduction in both research time and code costs compared to prior models. The company reports that the new model allows agents to complete complex knowledge work significantly faster while maintaining high-quality output.

arxiv arXiv cs.CL · 1d ago · 11 views

AgentRouter reduces agentic workflow costs by 72% via step-level model routing

Researchers propose AgentRouter, a lightweight classifier that optimizes multi-step agentic workflows by routing individual trajectory steps to appropriate model tiers rather than using a single frontier model for all tasks. The system addresses the inefficiency of enterprise systems that waste 60-80% of their inference budget on subtasks that smaller models can handle equally well.

arxiv arXiv cs.AI · 12d ago · 17 views

AgentZip reduces AI-agent sandbox memory by up to 8.7x

Researchers present AgentZip, a memory compression system designed specifically for high-fanout AI-agent sandboxes to address memory bottlenecks caused by concurrent sessions.