Qodo Aware Deep Research achieves 80% accuracy with a new coding benchmark, surpassing Openai's codex by 74%, and is the human race. Claude Code 64%, Google's GeminiCLI 45%
Qodo, an agent code quality platform, has announced Qodo Aware, the new flagship product of the enterprise platform. It features the industry's first deep search agent specifically designed to navigate enterprise-scale codebases. In benchmark tests, Qodo Aware deep search agents showed superior accuracy and speed compared to major AI coding agents when answering questions that require context from multiple repositories.
AI made the generated code simpler, but ensuring quality on a large scale has become even more difficult. Modern software systems span hundreds or thousands of interconnected code repositories, making it almost impossible for developers to have a comprehensive understanding of an organization's entire codebase. Although current AI coding tools are great for single repository tasks, they cannot traverse a complex web of dependencies and relationships. The 2025 AI Code Quality Report states that over 60% of developers have missed the context related to AI coding tools. Qodo Aware addresses this limitation with a context engine that runs deep research agents that can automatically navigate repository boundaries.
“Developers usually don't work on their own. They need to understand how changes in one service affect systems across their organization and how those systems have evolved to what they are now,” he said. Itamar Friedmanco-founder and CEO of Qodo. “Our deep search agents can analyze impacts, dependencies and historical contexts in seconds across thousands of files and hundreds of repositories. This can take hours or days for principal engineers. This eliminates the traditional speed quality trade-offs companies face when employing AI for development, and helps them understand why their code was built.
Also Read: Aithority Interview with Speakup CEO Tim Morrs
Qodo Aware features three different modes, each with an agent specialized for a different use case. Deep Research Agent performs comprehensive multi-step analysis across the repository, making it ideal for complex architectural questions and system-wide tasks. For faster code Q&A, the ASK agent provides a quick response through obtaining the agent context, and the problem finder agent searches the repository for bugs, code replication, security risks, and other hidden issues. These agents can be used to retrieve answers directly or integrated into cursors or existing coding agents. Claude Codea powerful context search layer that improves your ability to understand large codebases.
Qodo Aware uses a sophisticated indexing and context search approach that combines Language Server Protocol (LSP) analysis, knowledge graphs, and vector embedding to generate deep semantic understandings of code relationships. For businesses, this means that developers can safely modify complex systems without fear of breaking unknown dependencies, reducing deployment risks, and accelerating the release cycle. Teams report time to investigate complex issues from days to minutes, even when working beyond a large, interconnected codebase 100m Line of code.
In addition to these features, Qodo is releasing a new multireposit dataset to evaluate coding deep search agents. The dataset contains actual questions that require information spanning multiple open source code repositories to be answered correctly. In the new DeepCodeBench benchmark, Qodo Aware achieved 80% accuracy, while Openai Codex scored 74%. Claude Code It reached 64% and Gemini CLI resolved 45% correctly. Importantly, Qodo Aware Deep Research takes less than half the time of codex, allowing faster iteration cycles for developers.
Qodo Aware is directly integrated into existing Qodo development tools, such as the Qodo Gen IDE Agent, Qodo Command CLI Agent, and Qodo Merge Code Review Agent. Bring context to your workflows throughout the software development lifecycle. Qodo Aware can be deployed within an enterprise single tenant environment, maintaining the governance and compliance standards that enterprises need while keeping code from leaving organizational boundaries. Supports Github, Gitlab, and Bitbucket, all indexing and processing occurs within the customer-controlled infrastructure.
