TokenRhythm Raises New Round for Multi-Model Routing Infrastructure
The AMW Read
The round incrementally updates the AI infrastructure player map while highlighting model-routing infrastructure as a segment-level competitive layer for agent deployment.
TokenRhythm Raises New Round for Multi-Model Routing Infrastructure
TokenRhythm, a China-based AI infrastructure company, has completed a new financing round led by Honghui Fund, with Juhé Capital and Shangshi Capital participating. The company said its cumulative funding has reached tens of millions of dollars and its valuation is in the several-hundred-million-dollar range. Its routing platform automatically selects, combines, and switches among models based on task stage, quality requirements, and cost constraints.
The company is targeting the control layer between model providers and agent applications: a routing and scheduling system intended to make multi-model workflows cheaper and more reliable. TokenRhythm says its API platform supports OpenAI- and Claude-compatible protocols, has attracted 54,000 users, and processes more than 500 billion tokens daily. Its open-source agent, OpenSquilla, has surpassed 6,600 GitHub stars. These are company-reported operating metrics, not independently verified usage figures.
The strategic question is whether routing remains an interchangeable API feature or becomes a durable source of task-level data and workflow expertise. TokenRhythm argues that model-selection decisions, execution results, and user feedback can improve future routing and ultimately support models designed around real agent tasks. Builders should measure routing systems against workload-level outcomes such as completion quality, latency, failure recovery, and blended inference cost rather than headline model rankings. Investors should distinguish gateway volume from defensible orchestration data and enterprise integration depth.
