GitHub Updates: Transition to Gemini Pro and Upcoming Changes for Hardcoded Secrets

GitHub is continuously evolving, with new updates aimed at improving security and efficiency for developers. Two significant changes are making waves in the GitHub ecosystem: the transition from DeepSeek-V3 to DeepSeek-V3-0324 and the introduction of Gemini 2.5 Pro, the latest model for Copilot users. Additionally, GitHub is set to phase out CodeQL’s detection of hardcoded secrets in favor of a more precise and effective secret scanning feature. Here’s what you need to know to stay ahead of these changes.

GitHub’s Latest Update: Key Changes and Transitions

GitHub has officially deprecated DeepSeek-V3 in its GitHub Models. Developers are advised to transition to the improved version, DeepSeek-V3-0324, to continue enjoying enhanced features and updates. This transition ensures continued access to the most recent developments and optimizations.

In another major update, Google’s Gemini 2.5 Pro model is now available to all GitHub Copilot customers. This model, the latest iteration of Google’s AI for complex tasks, excels in reasoning, code generation, and problem-solving capabilities. It’s proven to outperform in math, science, and general coding benchmarks, offering Copilot users a more advanced tool for development tasks.

For Copilot Pro/Pro+ users, Gemini 2.5 Pro can now be accessed through the model selectors in Copilot Chat on VS Code and the immersive chat on GitHub. However, for Copilot Business or Enterprise users, administrators must enable this new model in Copilot settings through a new policy. After enabling, the model selector will appear in the interface for individual users.

In addition, GitHub is discontinuing the use of CodeQL for detecting hardcoded secrets starting on May 30, 2025. Developers are encouraged to transition to GitHub’s Secret Scanning feature, which offers more accurate detection of over 300 types of hardcoded secrets. This change aims to reduce redundant alerts and improve overall secret management.

What Undercode Says:

GitHub’s push for updates and improvements, such as the deprecation of DeepSeek-V3 and the introduction of Gemini 2.5 Pro, reflects the ongoing trend of pushing AI models to be more powerful, precise, and user-friendly. The Gemini 2.5 Pro model is particularly noteworthy as it brings advanced reasoning capabilities that were previously seen as niche to other specialized AI systems. For developers relying on GitHub Copilot, this improvement will undoubtedly increase productivity and ease of use, especially in complex coding tasks that require advanced AI assistance.

The decision to discontinue CodeQL’s hardcoded secrets detection in favor of Secret Scanning is another logical step. The redundancy of using both tools had created unnecessary complexity and manual overhead for developers. GitHub’s shift towards a more streamlined, efficient approach through Secret Scanning is expected to result in better accuracy, fewer false positives, and an overall reduction in the noise developers face when managing security alerts.

Secret scanning’s ability to detect over 300 types of hardcoded secrets makes it far superior in terms of coverage and precision. Integrating this feature with Copilot’s AI capabilities is a win for developers, allowing for a more cohesive workflow. The ability to manage all alerts in a single interface ensures that developers can focus on fixing vulnerabilities rather than managing multiple alert systems.

However, it is crucial for businesses and enterprise users to stay proactive in enabling these features. As the adoption of Gemini 2.5 Pro and the migration to Secret Scanning become standard, organizations should take the necessary steps to ensure they are aligned with GitHub’s updates. Failure to do so could lead to inefficiencies and missed opportunities to leverage the latest advancements in code scanning and security.

As GitHub continues to refine its tools, the broader implications for the development community are clear: a focus on improving AI, enhancing security, and streamlining workflows. Developers who adopt these new models and features early will likely find themselves at an advantage in terms of productivity and security, making these updates an essential part of staying competitive in the fast-paced world of software development.

Fact Checker Results:

The transition from CodeQL to Secret Scanning for hardcoded secrets is scheduled for May 30, 2025, and aligns with GitHub’s ongoing security improvements.
Gemini 2.5 Pro is indeed the latest and most advanced AI model for GitHub Copilot users, offering superior coding and problem-solving capabilities.
DeepSeek-V3 has been officially deprecated, and users must migrate to DeepSeek-V3-0324 to ensure continued access to enhanced features.