# Cisco Foundation AI Releases Antares: 350M and 1B Open-Weight Models That Localize Known Vulnerabilities Inside Real

URL: https://technosports.co.in/cisco-antares-models/  
Published: 2026-07-22  
Updated: 2026-07-22  
Author: Reetam Bodhak

Antares: On July 21, 2026, Cisco Foundation AI introduced its new suite, which consists of open-weight AI models aimed at tackling vulnerabilities in software security. This launch represents a significant step forward in how developers can spot and handle security flaws within actual codebases. The models are available in two sizes: 350 million parameters and 1 billion parameters, giving developers specialized tools to boost software security.

**Models focus on pinpointing vulnerabilities in codebases.**

![Cisco](https://technosports.co.in/wp-content/uploads/2026/07/ciscococ.jpg)

## Cisco : Key Details About Models

The models are specifically designed for one crucial task: vulnerability localization. This means they can take a vulnerability description and find the exact files in a code repository that contain the flaw. Both the -350M and -1B models are now available on Hugging Face and licensed under the Apache 2.0 license, making them widely usable across various software projects.

Cisco’s initiative also brings the Vulnerability Localization Benchmark (VLoc Bench) into the mix, featuring 500 tasks aimed at assessing the models’ performance.

In an interesting development, the -1B model achieved a File F1 score of 0.209. For comparison, the well-known GPT-5.5 scored 0.229, while a 753 billion parameter open-weight model scored 0.186. These results show that while this model hasn’t set a new benchmark yet, it still offers competitive capabilities in vulnerability localization.

However, Cisco points out that these models aren’t meant to replace existing application security tools. Instead, they aim to complement those tools by addressing the initial triage step, which often poses significant challenges and costs for developers trying to connect external vulnerability knowledge with internal source code. This knowledge typically lives in public databases, advisories, and Common Weakness Enumerations, while the code itself exists in complex, modular repositories. For more information, check out [VentureBeat AI](https://venturebeat.com/category/ai).

## Context of the Release

The development of these models highlights the growing awareness of the challenges developers face in managing software security. As codebases become more modular and filled with dependencies, the need for efficient vulnerability detection tools is more pressing than ever. These models tackle that need by simplifying the process of identifying flaws, potentially saving developers valuable time and resources.

The release of the models fits into a larger trend within the tech industry, where companies increasingly use AI to bolster cybersecurity measures. With the rise of sophisticated cyber threats, organizations must find inventive solutions to keep up with evolving vulnerabilities. Cisco’s focus on this particular task of vulnerability localization shows the value of specialized models in a field often characterized by general-purpose AI.

## What’s Next for Models and Developers

Right now, we don’t have specific details about the hardware specifications and system requirements for running these models. This lack of clarity leaves potential users with a few questions about implementation. However, since the models are open-weight, they can fit into various development environments, offering teams the flexibility to enhance their security protocols.

Looking ahead, these models could pave the way for more advanced AI-driven tools that further streamline vulnerability management. As AI and machine learning continue to advance, we can expect solutions to evolve, giving developers even more powerful capabilities to secure their software.

Cisco’s release has significant implications. Developers now have specialized tools that can effectively pinpoint vulnerabilities in their code, leading to more secure software applications. As the cybersecurity landscape shifts, tools like these could become essential for how organizations manage and mitigate risks.

---

## FAQs

### What are the sizes of the models?

The models are available in two sizes: 350 million parameters and 1 billion parameters.

### How do the models assist developers?

They localize known vulnerabilities within real codebases, helping developers identify security flaws more efficiently.

### Are the models open-weight?

Yes, both -350M and -1B are open-weight models licensed under Apache 2.0.

### What is the VLoc Bench?

The VLoc Bench is a 500-task benchmark designed to evaluate the performance of the models in vulnerability localization.

### What are the future implications of these models?

These models could lead to more advanced AI tools for managing vulnerabilities, ultimately enhancing software security for developers.

**Source:** [Marktechpost](https://www.marktechpost.com/2026/07/21/cisco-foundation-ai-releases-antares-350m-and-1b-open-weight-models-that-localize-known-vulnerabilities-inside-real-codebases/)