Jeffrey Quesnelle, CTO of Nous Research, said the development and impact of the Hermes AI platform during a recent interview on the "Intelligent Machines" podcast [1].
The rise of Hermes AI signals a shift in the open-source artificial intelligence landscape, demonstrating that community-driven agents can compete with established models in scale and utility.
Quesnelle, the creator of Hermes AI, used the appearance on the TWiT network to explain the capabilities of the agentic platform [1]. The project has seen rapid adoption among developers and users seeking open alternatives to proprietary AI systems.
Data indicates that Hermes achieved a significant milestone on May 10, 2026, when it overtook OpenClaw to become the top open-source AI agent on OpenRouter [2]. This transition marks a pivotal change in user preference for open-weight agentic frameworks.
The scale of this adoption is evident in the platform's processing volume. In a single day, Hermes processed 224 billion tokens [2]. During that same 24-hour period, OpenClaw processed 186 billion tokens [2].
These figures highlight the increasing demand for high-throughput, open-source agents that can handle massive datasets without relying on closed-ecosystem providers. Quesnelle said the goal of the project is to expand the accessibility and efficiency of agentic AI for the broader research community [1].
By providing a transparent alternative to commercial models, Nous Research aims to foster a more collaborative environment for AI development. The ability of Hermes to maintain such high token volumes suggests a robust architecture capable of supporting enterprise-level workloads, while remaining open to public scrutiny [2].
“Hermes processed 224 billion tokens in a single day”
The rapid ascent of Hermes AI over OpenClaw suggests a growing market preference for specific open-source agent architectures. When a single agent processes over 200 billion tokens in one day, it proves that open-source models are no longer just for experimentation but are capable of handling production-grade traffic at a scale previously reserved for proprietary giants.
