Claude Opus 4.7 Released by Anthropic

Anthropic has launched Claude Opus 4.7, enhancing its capabilities for complex tasks and introducing new features for improved user experience.

OpenAI has updated Codex and released the "reasoning" AI model GPT-Rosalind.

Anthropic has introduced Claude Opus 4.7, the most powerful model in the Opus series to date.

Introducing Claude Opus 4.7, our most capable Opus model yet.

It handles long-running tasks with more rigor, follows instructions more precisely, and verifies its own outputs before reporting back.

You can hand off your hardest work with less supervision. pic.twitter.com/PtlRdpQcG5
— Claude (@claudeai) April 16, 2026

The new version is available to all paid Claude users and via API at $5 per million input tokens and $25 per million output tokens.

Key Improvements

Opus 4.7 excels at complex tasks, with developers noting that users trust it with work that previously required close supervision.

In agent programming, the model outperformed its predecessor by 10%, and in visual data processing by 13%. Gains in other areas were more modest.

The model features significantly enhanced visual capabilities, processing images up to 2576 pixels on the long side (about 3.75 megapixels)—more than three times the capacity of previous Claude versions.

Source: Anthropic.

Opus 4.7 follows instructions more accurately. Prompts designed for older models may yield unexpected results, as they interpreted instructions loosely, while the new version takes them literally. Users are advised to adjust their queries accordingly.

Additionally, the latest Claude has learned to retain information between sessions—it saves notes in files and can reference them in future conversations.

Anthropic has added a new effort level, xhigh ("extra high"), positioned between high and max. This allows for more precise tuning of the balance between depth of analysis and response speed.

In Claude Code for all plans, the default effort level has been raised to xhigh.

Other new features include:

Task budgets (public beta API) — managing token expenditure;
/ultrareview — a dedicated code review session in Claude Code;
Auto mode for Max users — Claude makes decisions independently.

Cyber Capability Limitations

Opus 4.7 is weaker in cybersecurity compared to Mythos Preview. Anthropic intentionally reduced these capabilities during training. The model includes built-in protections that block prohibited and high-risk requests.

"What we learn from the real deployment of these protective mechanisms will help us move toward our ultimate goal — the broad release of Mythos-class models," noted the startup team.

Security professionals wishing to use Opus 4.7 for legitimate purposes (such as vulnerability research and pentesting) have been invited by Anthropic to join a new Cyber Verification program.

However, users have found these limitations problematic. Some clients report that the model refuses to write code, claiming it "detects malware in every request."

It suspects everything is malware, and still refuses to code after confirming that there is no malware pic.twitter.com/YXpaoNV8YG
— Hanh Nguyen (@fashiongiik) April 16, 2026

OpenAI's Response

OpenAI has released a "major update" to Codex, currently available only on macOS.

Codex for (almost) everything.

It can now use apps on your Mac, connect to more of your tools, create images, learn from previous actions, remember how you like to work, and take on ongoing and repeatable tasks. pic.twitter.com/UEEsYBDYfo
— OpenAI (@OpenAI) April 16, 2026

The new version can interact with user applications: it can see the screen, click, and type with its own cursor. On Mac, it supports multiple agents working in parallel without interfering with other software.

Built-in Browser, Plugins, and Development Lifecycle

Codex now includes a built-in browser: users can comment directly on pages, providing the agent with precise instructions. This feature is particularly useful for frontend and game development.

With computer use on macOS, Codex can now use any app by seeing, clicking, and typing with its own cursor.

It runs in the background without taking over your computer, working on tasks like frontend iteration, app testing, or any workflow that doesn't expose an API. pic.twitter.com/iO9iubLZX9
— OpenAI (@OpenAI) April 16, 2026

Developers plan to expand browser management capabilities beyond the local environment.

Codex also supports gpt-image-1.5 for generating and iterating images. Together with screenshots and code, this allows for creating visual concepts, frontend designs, mockups, and games within a single interface.

You can now generate and iterate on images with gpt-image-1.5 in Codex to create frontend designs, mockups, game assets, and more without leaving your workflow.

Usage is included with your ChatGPT account, no API key needed. pic.twitter.com/ay17I3Nxoa
— OpenAI (@OpenAI) April 16, 2026

OpenAI has released over 90 additional plugins that combine skills, application integrations, and MCP servers. These include Atlassian Rovo for working with JIRA, CircleCI, CodeRabbit, GitLab Issues, Microsoft Suite, Neon by Databricks, Remotion, Render, and Superpowers.

Codex now supports GitHub comments, multiple terminal tabs, and remote devbox connections via SSH (in alpha).

Users can open files directly in the sidebar with enhanced previews for PDFs, spreadsheets, slides, and documents, as well as utilize a new summary panel to track agent plans, sources, and artifacts.

Memory and Planning

Codex has gained the ability to plan future work and automatically resume long-term tasks—potentially spanning days or weeks. Teams use automation for various purposes, from code review requests to task tracking in Slack, Gmail, and Notion.

Source: OpenAI.

Developers have improved the assistant's memory. Codex can now remember useful context from past dialogues—personal preferences and corrections.

The model also actively suggests helpful actions, continuing from where the user left off. For example, the agent can find open comments in Google Docs, extract context from Slack, Notion, and the codebase, and then provide a prioritized list of actions.

New GPT Model

Additionally, OpenAI has introduced the "reasoning" AI model GPT-Rosalind to accelerate drug development.

Introducing GPT-Rosalind, our frontier reasoning model built to support research across biology, drug discovery, and translational medicine. pic.twitter.com/PubLU0FkSv
— OpenAI (@OpenAI) April 16, 2026

It is named after English biophysicist Rosalind Franklin, whose research helped uncover the structure of DNA and laid the groundwork for modern molecular biology.

Developers noted that in the U.S., developing a new drug typically takes 10-15 years. The future of a drug is largely determined in the early stages of research. The greatest challenges involve analyzing vast amounts of scientific publications and specialized databases.

The goal of GPT-Rosalind is to serve as a biologist's assistant: summarizing scientific texts, formulating hypotheses, designing experimental plans, and processing information. The model is particularly strong in tasks related to proteins, molecules, genes, and similar biological structures.

In the BixBench benchmark (real bioinformatics analysis), GPT-Rosalind achieved one of the best results among models with published data.

In LABBench2, it outperformed GPT-5.4 in six out of 11 tasks. The most notable lead was in CloningQA, which requires designing DNA and enzymes for molecular cloning protocols.