OpenAI expands Codex and GPT‑Rosalind
OpenAI rolled out a life‑sciences model called GPT‑Rosalind and broadened Codex's capabilities and plugin ecosystem, bringing more data‑analysis and bioinformatics tooling into limited access releases. Reports say Codex is moving beyond simple code completion toward controlling computers, scheduling work and running long‑running automations via new plugins and in‑app browser features. (venturebeat.com) (help.openai.com) (iclarified.com)
OpenAI on April 16 expanded Codex beyond code generation and introduced GPT‑Rosalind, a new life‑sciences model in limited release. (openai.com 1) (openai.com 2) Codex’s new desktop app for macOS and Windows adds computer use, an in‑app browser, image generation, memory, and plugins, according to OpenAI’s product post published Thursday. OpenAI said more than 3 million developers already use Codex each week. (openai.com) OpenAI said Codex can now “see, click, and type” on a user’s computer with its own cursor, run multiple agents in parallel on a Mac, connect to remote development boxes over Secure Shell, review pull requests, and work across multiple files and terminals. The in‑app browser lets users comment directly on pages while an agent iterates on frontend work. (openai.com 1) (openai.com 2) The GitHub release log for OpenAI’s codex repository shows the tooling underneath that shift: version 0.121.0 added `marketplace add` and app‑server support for installing plugin marketplaces from GitHub, git URLs, local directories, and direct `marketplace.json` links. The same release expanded Model Context Protocol, or MCP, support with namespaced registration and MCP app tool calls. (github.com) OpenAI’s separate plugins repository now highlights bundles for Figma, Notion, iOS app building, macOS app building, web app deployment, Netlify, Remotion, and Google Slides. That turns Codex into more of a hub for connected developer tools than a standalone autocomplete system. (github.com) GPT‑Rosalind is aimed at a different problem: early drug discovery and biology research, where teams sift through papers, databases, lab results, and competing hypotheses before they ever run a clinical trial. OpenAI said the model is optimized for workflows across chemistry, protein engineering, and genomics. (openai.com) OpenAI said GPT‑Rosalind is available as a research preview in ChatGPT, Codex, and the application programming interface for qualified customers through its trusted access program. The company also launched a free Life Sciences research plugin for Codex that connects to more than 50 scientific tools and data sources. (openai.com) The help center says GPT‑Rosalind is currently limited to eligible U.S. customers with Enterprise agreements, and individual researchers are not supported. OpenAI said the model is for internal research workflows, not customer‑facing products or external commercial applications during the preview. (help.openai.com) OpenAI said the model is most useful for target biology, mechanism research, literature synthesis, and omics interpretation, and named Amgen, Moderna, the Allen Institute, and Thermo Fisher Scientific among early customers. The company said the model is named for Rosalind Franklin, whose work helped reveal DNA’s structure. (openai.com) Taken together, the April 16 releases push OpenAI’s product line in two directions at once: Codex toward longer‑running software agents, and GPT‑Rosalind toward specialized scientific work inside enterprise labs. Both are still gated behind waitlists, previews, or enterprise access controls rather than broad public release. (openai.com) (help.openai.com)
