Cisco (CSCO) has made a surprising leap in AI-driven vulnerability detection, unveiling a family of small language models (SLMs) called Antares that it claims outperforms OpenAI‘s GPT-5.5 and Google‘s Gemma and Gemini models. However, the company has notably omitted any comparison with Anthropic‘s latest and most powerful model, Mythos, leaving the industry guessing about where it truly stands against its Project Glasswing partner’s top-tier AI.
What Happened: Cisco’s Antares SLMs Enter the Ring
Cisco released two Antares models — with 350 million and 1 billion parameters — on Hugging Face as open-weight releases, allowing developers to download the model files without full training data or code. According to Amin Karbasi, Cisco’s VP and chief AI scientist, the models mimic how a “human investigator works through a repository,” starting from a vulnerability description, searching for relevant code patterns, reading candidate files, and narrowing down to the most likely vulnerable files. Antares is specialized for repository-level vulnerability localization, supporting workflows such as locating files by Common Weakness Enumeration (CWE) category, triaging security advisories, augmenting static analysis, and enabling shift-left security in CI/CD pipelines.
Why It Matters: The Small Model Rebellion in Security
Research by the Cisco Foundation AI team showed Antares scoring higher than OpenAI’s GPT-5.5 and Google’s Gemma and Gemini models at vulnerability finding — though it did not surpass OpenAI’s premium GPT-5.5 (xhigh) model. Interestingly, a technical report from the same period revealed that Anthropic‘s Claude Opus 4.6 tested higher than both Antares and GPT-5.5 (xhigh) on benchmarks. Yet these results were absent from Cisco’s Antares announcement, presumably because the Anthropic tests were run through Claude Code without identical parameters. Cisco is a member of Anthropic’s Project Glasswing, which grants exclusive access to Mythos, but Mythos was not mentioned in any testing materials. This omission has fueled speculation: does Mythos vastly outperform Antares, or are contractual restrictions at play?
Our Analysis: A Paradigm Shift in the AI Security Market
XPLAIN AI interprets this event as the beginning of a “small model rebellion” in cybersecurity. Specialized SLMs are proving that they can outperform general-purpose large language models (LLMs) on domain-specific tasks. Cisco, with its dominant position in enterprise security, can integrate Antares into its existing product suite and leverage its massive customer base to rapidly gain market share. This poses a direct threat to competitors like CrowdStrike (CRWD), Palo Alto Networks (PANW), and Zscaler (ZS), which may struggle to match Cisco’s AI capabilities. On the flip side, NVIDIA (NVDA) could benefit from increased demand for inference GPUs as SLM adoption grows. However, the relationship with Anthropic is a double-edged sword: while Project Glasswing provides access to cutting-edge technology, the lack of Mythos comparison could be interpreted negatively by the market, casting doubt on Cisco’s AI prowess.
Stakeholder Impact: Winners and Risks
- Potential beneficiaries: Cisco (CSCO) could retain customers and generate new revenue by embedding Antares into its security products. NVIDIA (NVDA) may see indirect gains from increased SLM inference workloads. Open-source AI advocates like Meta (META) with its Llama models could also benefit from the validation of smaller, specialized models.
- Risks: CrowdStrike (CRWD), Palo Alto Networks (PANW), and Zscaler (ZS) face competitive pressure if Cisco’s AI proves superior. The undisclosed Mythos comparison may weigh on Cisco’s stock in the short term, as investors question the true state of its AI technology.
Counter-Scenarios and Uncertainties
Several factors could challenge this narrative. First, Cisco’s benchmark results may not replicate in real-world environments, where noise and complexity differ. Second, if Anthropic’s Mythos is significantly more capable, Cisco’s AI strategy may remain that of a follower rather than a leader. Third, OpenAI and Google could quickly release security-specific models that outperform Antares. Investors should monitor Cisco’s quarterly security product revenue growth, customer adoption rates, and any updates from Anthropic regarding Mythos’s public availability.
#Cisco #AISecurity #Antares #SLM #VulnerabilityDetection #Anthropic #Mythos #OpenAI #Google #Cybersecurity #NVIDIA
Sources
- Cisco security AI aces OpenAI – but Anthropic? That's a Mythos mystery — SDxCentral · News coverage · Tue, 21 Jul 2026 15:30:12 +0000
Written by: XPLAIN AI Editorial Team · Reviewed by: XPLAIN AI Editorial Desk
This content was drafted with AI assistance based on publicly available sources and reviewed under XPLAIN AI's editorial standards.