entry 3.5 Flash dominates the OpenClaw ranking for three consecutive days

Just over two weeks ago, OpenRouter—an API aggregation platform for large language models—revealed a notable fact: Stepfun’s entry 3.5 Flash became the preferred model on OpenClaw, the open-source AI assistant framework, holding the number one spot for three consecutive days from March 4 to 6.

Historic Record: 61.3 Billion Tokens Consumed

The most remarkable data point came on March 5, when entry 3.5 Flash reached an unprecedented volume of 61.3 billion invocations in a single day. This figure not only reflects the model’s popularity among OpenClaw developers but also its efficiency in processing large-scale queries. The ability to generate and manage this massive amount of tokens within 24 hours positions 3.5 Flash as a robust option for high-performance applications.

Global Podium: Asian AI Dominance

Completing the top spot in the ranking, other top-tier models competed for the following positions. MiniMax M2.5, Trinity Large Preview from Arcee AI, Kimi K2.5 from Moonshot, and Claude Sonnet 4.6 from Anthropic were the closest competitors during that period.

The most notable aspect of this ranking is the prominence of Chinese models: three of the top five spots were occupied by products developed in Asia, demonstrating the technological advancement and competitiveness these models have gained in the global AI market. This performance indicates a significant shift in the balance of power among the leading providers of language models worldwide.