Home
Cisco opens source Antares AI to help enterprises accurately identify code vulnerabilities

Cisco has officially open-sourced its new AI model family, Antares, to bolster cybersecurity defense. Released on Tuesday, this lightweight AI suite is engineered specifically for analyzing software code repositories, enabling the precise identification of known security vulnerabilities within source code.
Unlike large language models that demand substantial computational resources, Antares delivers high accuracy with remarkable efficiency. Benchmarks indicate the model can scan 500 code repositories in just 15 minutes, incurring costs of less than $1.
Enhancing local privacy while leveraging hacker methodologies for precise searches
A primary benefit of this compact model is its ability to run locally within enterprise environments, thereby eliminating the need to transmit sensitive source code to external third-party services. Its training framework mirrors that of seasoned security investigators, allowing it to autonomously trace potential threats across extensive codebases.
To mitigate the risk of misuse, Cisco has partnered with industry organizations to implement a download verification process. Currently, the Antares-350M and 1B versions are available on open-source platforms exclusively for verified security professionals.
Strengthening the open-source security ecosystem through industry collaboration
The more powerful 3B version remains unreleased to the public; Cisco intends to integrate it directly into its proprietary security product lineup. Additionally, the company is exploring the formation of an industry alliance to expand cooperation in the realm of open-source AI security tools.
As numerous companies introduce similar solutions, collaboration within the open-source security sector is intensifying. This efficient, cost-effective AI protection tool is poised to establish a more robust defense for cybersecurity in the digital era.
Related article
DeepMind CEO Urges US-Led Testing of AI Models
Demis Hassabis, Co-Founder and CEO of Google DeepMind. Credit: Demis Hassabis/LinkedInGoogle DeepMind CEO Demis Hassabis highlights the dangers of artificial general intelligence and advocates for a US-led governance frameworkGoogle DeepMind’s CEO, D
OpenAI Launches Linux ChatGPT App to Complete Desktop OS Support
OpenAI launched a preview of its ChatGPT desktop app for Linux on August 11, 2026, bringing ChatGPT and Codex tools to Linux users and completing support for all major desktop operating systems.This move addresses long-standing requests from open-sou
Alipay Cloud Developer WeChat Account Renamed Qwen AI Platform
On July 22, the original WeChat official account "Aliyun Developer" was officially renamed to "Qwen AI Platform". This brand upgrade aims to adapt to the new trend of AI evolving from an auxiliary tool to an independent thinking, collaborating, and c
Related Special Topic Recommendations
Comments (0)
0/500

Cisco has officially open-sourced its new AI model family, Antares, to bolster cybersecurity defense. Released on Tuesday, this lightweight AI suite is engineered specifically for analyzing software code repositories, enabling the precise identification of known security vulnerabilities within source code.
Unlike large language models that demand substantial computational resources, Antares delivers high accuracy with remarkable efficiency. Benchmarks indicate the model can scan 500 code repositories in just 15 minutes, incurring costs of less than $1.
Enhancing local privacy while leveraging hacker methodologies for precise searches
A primary benefit of this compact model is its ability to run locally within enterprise environments, thereby eliminating the need to transmit sensitive source code to external third-party services. Its training framework mirrors that of seasoned security investigators, allowing it to autonomously trace potential threats across extensive codebases.
To mitigate the risk of misuse, Cisco has partnered with industry organizations to implement a download verification process. Currently, the Antares-350M and 1B versions are available on open-source platforms exclusively for verified security professionals.
Strengthening the open-source security ecosystem through industry collaboration
The more powerful 3B version remains unreleased to the public; Cisco intends to integrate it directly into its proprietary security product lineup. Additionally, the company is exploring the formation of an industry alliance to expand cooperation in the realm of open-source AI security tools.
As numerous companies introduce similar solutions, collaboration within the open-source security sector is intensifying. This efficient, cost-effective AI protection tool is poised to establish a more robust defense for cybersecurity in the digital era.
DeepMind CEO Urges US-Led Testing of AI Models
Demis Hassabis, Co-Founder and CEO of Google DeepMind. Credit: Demis Hassabis/LinkedInGoogle DeepMind CEO Demis Hassabis highlights the dangers of artificial general intelligence and advocates for a US-led governance frameworkGoogle DeepMind’s CEO, D
OpenAI Launches Linux ChatGPT App to Complete Desktop OS Support
OpenAI launched a preview of its ChatGPT desktop app for Linux on August 11, 2026, bringing ChatGPT and Codex tools to Linux users and completing support for all major desktop operating systems.This move addresses long-standing requests from open-sou
Alipay Cloud Developer WeChat Account Renamed Qwen AI Platform
On July 22, the original WeChat official account "Aliyun Developer" was officially renamed to "Qwen AI Platform". This brand upgrade aims to adapt to the new trend of AI evolving from an auxiliary tool to an independent thinking, collaborating, and c











