Nous Research Launched Agentic AI Benchmarking Index

The firm introduced a standardized scoring system for autonomous agents alongside a $90 million funding round.

Updated on Oct. 7, 2026 in Artificial Intelligence

Isometric editorial illustration of server racks representing data center infrastructure for AI model benchmarking. — Nous Research launched its Hermes Index on Tuesday, providing a standardized framework for measuring the performance and cost-efficiency of frontier agentic AI models. AI Illustration. Upload story photo >

Live Poll

Do you prioritize benchmark scores when choosing AI tools for your work or business?

Nous Research has launched the Hermes Index, a new benchmarking framework designed to evaluate the performance and cost-efficiency of 14 frontier agentic AI models. The release accompanies a $90 million funding round that brings the company's total capital to $160 million.

Why it matters

As the industry shifts toward autonomous agents, establishing standardized metrics for task completion and cost is critical for scaling enterprise deployments. Nous Research intends to utilize its $1.5 billion valuation and new capital to accelerate these enterprise-grade capabilities.

The Hermes Index evaluates models across the Hermes Bench, Terminal-Bench 4.0, and SkillsBench suites. Claude Opus 5.5 achieved a top score of 63.31, while smaller models like DeepSeek V4.1 Flash achieved a 36.91 score at a significantly lower cost of $0.259 per task.

The players

Nous Research

An AI firm focused on developing agentic models and evaluation frameworks, now valued at $1.5 billion.

Nvidia

A leading provider of high-performance GPUs and hardware infrastructure supporting large-scale AI research.

M12

The venture capital arm of Microsoft that invests in enterprise-focused software and AI technology stacks.

The details

The index measures model performance by executing 150 tasks that require autonomous reasoning and tool use. These evaluations are processed through an agent evaluation harness—a software tool designed to manage and record the steps taken by an AI model to reach a final output. By aggregating scores across four distinct test suites, the platform provides a quantitative comparison of task accuracy versus the associated inference costs for each tested architecture.

2023: Nous Research was founded.
February 25, 2026: The company started developing the Hermes Agent evaluation harness.
October 6, 2026: The Hermes Index was officially launched.
October 7, 2026: The company secured $90 million in new funding.

The Tech Race

The Hermes Index provides a standardized benchmark for agentic models, following the industry trend of shifting focus from static language performance to functional task execution. This launch marks a formal effort to categorize the capabilities of frontier models within a crowded, competitive market.

Enterprises can use these benchmarks to select models that balance performance requirements with specific task budgets. While the index is currently available for comparative analysis, actual deployment depends on individual model availability and internal API integration.

The takeaway

The emergence of the Hermes Index reflects an industry-wide need for functional performance data as models transition from chatbots to active agents. Watch for updates to the benchmark scorecards as frontier developers release new iterations following the current round of capital infusions.