# Arm sets out an agentic CPU method and AGI CPU benchmark results

> Signal65 used server-level measurements to model rack-level results across seven workloads; Arm reported a 1.99x modeled average for AGI CPU.

- Canonical URL: https://newsverum.com/en/news/arm-agi-cpu-agentic-benchmark-methodology-october-2026
- Publisher: Newsverum
- Section: Technology
- Published: 2026-10-13T10:04:09+00:00
- Updated: 2026-10-13T10:04:09+00:00
- Event / data period: 2026-10-13
- Other editions: [Türkçe](https://newsverum.com/tr/news/arm-agi-cpu-agentic-benchmark-methodology-october-2026.md), [العربية](https://newsverum.com/ar/news/arm-agi-cpu-agentic-benchmark-methodology-october-2026.md), [Français](https://newsverum.com/fr/news/arm-agi-cpu-agentic-benchmark-methodology-october-2026.md), [Español](https://newsverum.com/es/news/arm-agi-cpu-agentic-benchmark-methodology-october-2026.md), [Deutsch](https://newsverum.com/de/news/arm-agi-cpu-agentic-benchmark-methodology-october-2026.md), [Português](https://newsverum.com/pt/news/arm-agi-cpu-agentic-benchmark-methodology-october-2026.md)

Arm introduced an agent-focused CPU evaluation method on October 13, with support from Fujitsu, Meta and OpenAI. The company said the same benchmark scripts ran on Arm AGI CPU and AMD EPYC servers, and that Signal65 independently verified the server-level measurements. The framework is intended to combine representative agent use cases, repeatable CPU measurements and checks against end-to-end outcomes.

The rack comparison Arm disclosed is a model derived from measured server performance, rather than a direct measurement of a complete rack. In that model, AGI CPU averaged 1.99 times the rack-level performance of EPYC 9755 across seven workloads; category averages were 1.7 for tooling, 2.0 for agent workflows and orchestration, and 2.2 for reasoning and planning. Arm separately reported up to 45% lower sandbox-creation latency against EPYC 9965, a different comparison system. These findings are limited to the workloads, server measurements and rack assumptions described by Arm.

## Data

### Workload count in Arm’s benchmark

Number of workloads compared in the rack-level modeling.

Methodology: Arm’s announcement reports seven workloads.

| Measure | Value |
|---|---|
| Arm rack-model workloads | 7 |

Sources: <https://newsroom.arm.com/news/arm-agi-cpu-agentic-benchmarks> · Download data: [JSON](https://newsverum.com/api/v1/articles/arm-agi-cpu-agentic-benchmark-methodology-october-2026/data/workload-count?lang=en&format=json) · [CSV](https://newsverum.com/api/v1/articles/arm-agi-cpu-agentic-benchmark-methodology-october-2026/data/workload-count?lang=en&format=csv)

### Sandbox latency result reported by Arm

Upper-bound latency reduction in Arm’s announcement.

Methodology: 45% is an upper bound against EPYC 9965; EPYC 9755 is the separate rack-performance comparator.

| Comparison | Reduction |
|---|---|
| Sandbox creation latency reduction versus EPYC 9965 (upper bound) | 45 |

Sources: <https://newsroom.arm.com/news/arm-agi-cpu-agentic-benchmarks> · Download data: [JSON](https://newsverum.com/api/v1/articles/arm-agi-cpu-agentic-benchmark-methodology-october-2026/data/sandbox-latency-reduction?lang=en&format=json) · [CSV](https://newsverum.com/api/v1/articles/arm-agi-cpu-agentic-benchmark-methodology-october-2026/data/sandbox-latency-reduction?lang=en&format=csv)

## Sources

- [Arm Newsroom](<https://newsroom.arm.com/news/arm-agi-cpu-agentic-benchmarks>) (Source checked: 2026-10-13; Published: 2026-10-13)

---

Cite the canonical URL above. [Editorial standards and corrections](https://newsverum.com/en/editorial)