---
type: "post"
id: "open-weight-models-trail-on-cybersecurity-eeeb2c"
slug: "open-weight-models-trail-on-cybersecurity-eeeb2c"
title: "Open-weight models trail on cybersecurity."
description: "Epoch AI published a comparison of open- and closed-weight AI models on Cyber ECI, its cybersecurity-focused version of the Epoch Capabilities Index."
retrieval_nugget: "Epoch AI published a comparison of open- and closed-weight AI models on Cyber ECI, its cybersecurity-focused version of the Epoch Capabilities Index."
published_at: "2026-10-24"
updated_at: "2026-10-24"
record_date: "2026-10-24"
date_kind: "published_at"
topics: ["claude"]
entities: ["Epoch AI","Cyber ECI","Epoch Capabilities Index","GLM-5.3","Z.ai","Claude Opus 5.5","GPT-6 Astra","OpenAI"]
related_patterns: ["ai-engineering-evaluations-are-splitting-into-task-specific"]
novelty_label: "new_fact"
what_changed: "Epoch’s Cyber ECI analysis scores GLM-5.3 at 160 and estimates the open cyber frontier trails closed models by 6–7 months."
prior_refs: ["memory:e3fdd306-5796-8b32-a320-faa3811eb6f8"]
source_url: "https://epoch.ai/data-insights/open-closed-cyber-eci-gap"
source_urls: ["https://epoch.ai/data-insights/open-closed-cyber-eci-gap"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:open-weight-models-trail-on-cybersecurity-eeeb2c"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/open-weight-models-trail-on-cybersecurity-eeeb2c/","markdown":"https://newruntime.com/posts/open-weight-models-trail-on-cybersecurity-eeeb2c.md","json":"https://newruntime.com/posts/open-weight-models-trail-on-cybersecurity-eeeb2c.json"}
---

# Open-weight models trail on cybersecurity.

## Retrieval answer

Epoch AI published a comparison of open- and closed-weight AI models on Cyber ECI, its cybersecurity-focused version of the Epoch Capabilities Index.

Epoch AI published a comparison of open- and closed-weight AI models on Cyber ECI, its cybersecurity-focused version of the Epoch Capabilities Index. The analysis says the open-weight frontier trails the closed-weight frontier by roughly six to seven months.

Z.ai's GLM-5.3 is listed as the highest-scoring open-weight model, with a Cyber ECI score of 160.1 and a 90% interval of 155.83 to 166.97. Epoch lists Claude Opus 5.5 at 171.14 and OpenAI's GPT-6 Astra at 171.03 among the current closed-weight leaders. The page classifies GLM-5.3 as restricted-use open weights; the two leading closed models are available by API.

The index is derived from cybersecurity benchmark results and a statistical fit shared with Epoch's broader ECI methodology. It does not establish real-world exploit success rates, deployment safety, or the security consequences of releasing any particular model's weights.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Claude](https://newruntime.com/topics/claude.md): Explore the claude topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)