Anthropic has launched Claude Opus 4.7, enhancing its AI model with improved capabilities for complex tasks and visual data processing.
Anthropic has unveiled Claude Opus 4.7, marking the most powerful iteration in its Opus series to date. This latest model is designed to handle complex tasks with improved precision and efficiency.
Claude Opus 4.7 is our most capable model yet, adept at long-running tasks, following instructions accurately, and verifying its outputs before reporting.
— Claude (@claudeai) April 16, 2026
Available to all paid users and via API, the pricing is set at $5 per million input tokens and $25 per million output tokens.
Key Enhancements
The model has shown a 10% improvement in agent programming tasks and a 13% boost in visual data processing capabilities compared to its predecessor. Claude Opus 4.7 now supports image processing up to 2576 pixels on the long side, significantly enhancing its visual capabilities.
One notable upgrade is its ability to retain information across sessions, allowing it to save notes and reference them in future interactions. Additionally, a new effort level termed "xhigh" has been introduced, providing users with finer control over the balance of analysis depth and response speed.
Other features include:
- Task budgets in public beta for managing token usage;
- /ultrareview for dedicated code review sessions;
- Auto mode for Max users, enabling the model to make independent decisions.
Cybersecurity Limitations
Despite its advancements, Opus 4.7 exhibits reduced capabilities in cybersecurity compared to the Mythos Preview, as Anthropic has intentionally scaled back these features during training. Built-in protections are designed to block high-risk requests, although some users have expressed frustration, noting that the model often misidentifies benign code as malware.
It suspects everything is malware, and still refuses to code after confirming that there is no malware.
— Hanh Nguyen (@fashiongiik) April 16, 2026
Anthropic is inviting security professionals to participate in a new Cyber Verification program for legitimate applications.
OpenAI's Competitive Response
In parallel, OpenAI has released a significant update to Codex, enhancing its functionality on macOS. The updated model can now interact with user applications, perform tasks independently, and includes a built-in browser for direct page interaction.
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.
— OpenAI (@OpenAI) April 16, 2026
Codex also supports the generation of images and has introduced over 90 new plugins for enhanced integration and functionality.
Introducing GPT-Rosalind
OpenAI has launched GPT-Rosalind, a new reasoning AI model aimed at accelerating drug development. Named after biophysicist Rosalind Franklin, the model assists researchers by summarizing scientific texts and designing experimental plans.
GPT-Rosalind is built to support research across biology, drug discovery, and translational medicine.
— OpenAI (@OpenAI) April 16, 2026
In benchmarks, GPT-Rosalind demonstrated strong performance, especially in bioinformatics tasks, outperforming GPT-5.4 in several categories.
Additionally, OpenAI has released a free Life Sciences plugin for Codex, enabling connections to over 50 public scientific databases.
On April 16, Google also launched Gemini 3.1 Flash TTS, an updated speech synthesis model.
