OpenAI Launches GPT-Rosalind for Drug Discovery and Updates Codex with Agentic Features

OpenAI has launched GPT-Rosalind, a specialized AI model with limited access designed to accelerate drug discovery in life sciences, alongside a major upgrade to its Codex tool, now available as a broader plugin on GitHub and enhanced desktop apps for developers. These releases position the company to tackle lengthy drug development timelines—often 10 to 15 years and billions of dollars—by streamlining fragmented workflows that bog down researchers, according to reports from VentureBeat.

The GPT-Rosalind model targets the grueling process from lab hypothesis to pharmacy shelf, where biology's complexities and manual shifts between experimental tools, software, and databases slow progress. Bloomberg notes that this early version pits OpenAI directly against Google in a competitive field, as tech giants race to demonstrate AI's potential for scientific breakthroughs like faster drug discoveries. Access remains restricted, likely to select partners or researchers, emphasizing controlled deployment in a high-stakes sector.

At the same time, OpenAI is supercharging Codex, its agentic coding tool, with dramatic updates that extend its reach across users' desktops. VentureBeat and TechCrunch describe how the revamped Codex desktop apps for Mac and Windows can now interact with other programs—operating apps with user permission, generating images via gpt-image-1.5, previewing webpages, and even prototyping websites for inline feedback. This "Super App" evolution includes over 100 new plugins, multiple terminal tabs, GitHub PR reviews, and a summary pane for tracking plans and artifacts, bringing it closer to a comprehensive developer environment with 3 million weekly users.

These moves reflect OpenAI's sharpening focus on business and enterprise tools amid intensifying rivalry. Fast Company reports that executives, including CFO Sarah Friar, are pivoting toward productivity aids like email and Slack summarization for work, while scaling back some consumer features to counter pressure from competitors like Anthropic. TechCrunch highlights Codex's beefed-up capabilities as a direct shot at Anthropic, granting more desktop control to automate coding, debugging, and shipping software.

For developers and creators, the updates promise tighter workflows. The built-in browser and computer use features—requiring screen recording and accessibility permissions on macOS—enable rapid frontend iteration and app interactions, as detailed in OpenAI's developer docs. This extends beyond coding to broader productivity, such as creating presentations or product mockups, while maintaining safeguards like app-specific permissions.

The releases come as OpenAI navigates internal and external challenges. A high-profile legal battle looms, with Elon Musk's lawsuit against CEO Sam Altman set for jury trial, questioning if the company has deviated from its original mission to benefit humanity through AGI, as covered by Wired. Meanwhile, tools like these could empower niches such as Roblox creators, where similar agentic AI aids game planning, building, and testing.

Business users, pharmaceutical researchers, and the 3 million-strong developer community stand to gain most, potentially slashing development times and costs. What happens next includes wider rollout—Codex now spans Windows and macOS globally, barring some regions like the EEA—along with monitoring for adoption and any regulatory scrutiny in sensitive areas like drug discovery. OpenAI's push underscores AI's shift from chatbots to practical agents reshaping industries.