Anthropic Launches Advanced AI Model Amid Safety Focus
SAN FRANCISCO — Anthropic has released Claude Opus 4.7, its latest AI model, on platforms including GitHub Copilot and Amazon Bedrock, company officials announced. The update brings significant improvements in coding, workflows and knowledge tasks, with benchmarks showing it resolves three times more production tasks than its predecessor.
The launch, which included a GitHub rollout on April 16, 2026, positions Opus 4.7 as a step toward safer AI deployment, according to GitHub Changelog and Anthropic statements. It features built-in safeguards to block high-risk cybersecurity requests, serving as a test for an unreleased "Mythos-class" model that remains internal due to potential risks, Anthropic and YouTube analysis indicated.
These developments expand Anthropic's ecosystem, including the unveiling of Claude Design for visual collaboration and participation in Project Glasswing, a multi-company effort to secure software. Amid intensifying AI competition, the moves emphasize enterprise utility and security.
Benchmark Gains in Coding and Workflows
Anthropic upgraded Claude Opus 4.7 to manage complex engineering tasks with minimal supervision, users reported. The model excels in agentic workflows, vision capabilities and multi-step processes, according to Anthropic and AWS.
Key benchmarks underscore the advances:
- 64.3% on SWE-bench Pro, an improvement over prior versions.
- 87.6% on SWE-bench Verified.
- 69.4% on Terminal-Bench 2.0.
- 64.4% on Finance Agent v1.1.
These scores demonstrate stronger performance in long-horizon reasoning, systems engineering and financial analysis, Anthropic said. Availability on GitHub Copilot and Amazon Bedrock includes a next-gen inference engine, public beta task budgets for token control and an updated tokenizer offering 1.0-1.3 times more tokens per input, per GitHub Changelog and AWS.
The update addresses feedback on code drift in multi-step tasks from Opus 4.6, supporting async workflows, CI/CD integration and self-verification for handling ambiguity, according to Anthropic, GitHub Changelog and YouTube analysis. "Users report being able to hand off their hardest coding work—the kind that previously needed close supervision—to Opus 4.7 with confidence," Anthropic stated in its announcement.
Embedded Safeguards and Ad-Free Commitment
Anthropic integrated safeguards into Opus 4.7 to automatically detect and block prohibited cybersecurity requests, officials explained. This prepares for future releases of more powerful models, including the internal "Mythos-class" Claude Mythos Preview, which remains unreleased due to risks, according to Anthropic and YouTube analysis.
"We are releasing Opus 4.7 with safeguards that automatically detect and block requests that indicate prohibited or high-risk cybersecurity uses," Anthropic said. The company also reaffirmed its ad-free approach for Claude, arguing that advertising conflicts with building a genuinely helpful AI assistant.
"Claude will remain ad-free. We explain why advertising incentives are incompatible with a genuinely helpful AI assistant," Anthropic noted. A large user study with 81,000 Claude.ai participants revealed insights into desires, dreams and fears about AI, though full details are limited in public reports, Anthropic said.
Expanding Tools and Security Alliances
Anthropic Labs introduced Claude Design, a tool for collaborative visual work such as designs, prototypes, slides and one-pagers. Powered by Claude models, it extends AI into creative workflows, company officials said.
Anthropic joined Project Glasswing, uniting firms including Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorgan Chase, Linux Foundation, Microsoft, NVIDIA and Palo Alto Networks to secure critical software, according to Anthropic. The initiative focuses on infrastructure security amid rising AI risks, echoing open-source efforts like those from the Linux Foundation.
No direct statements from partners like Cisco appeared in available sources, with secondary outlets such as Cisco Newsroom showing unrelated content.
Enterprise Impact and Future Directions
Opus 4.7 targets enterprise needs, accelerating development in fintech and software while shifting AI from chatbots to production tools, AWS and Anthropic said. Its multimodal capabilities align with trends in agentic AI and vision tasks, competing with tools like Adobe Firefly and rivals from OpenAI and Google, according to Anthropic and YouTube analysis.
Project Glasswing highlights growing collaboration among tech giants on security, potentially shaping hyperscaler standards, per Anthropic. Customer feedback has been positive: "Anthropic has already set the standard for coding models, and Claude Opus 4.7 pushes that further in a meaningful way as the state-of-the-art model on the market," one customer said, according to Anthropic.
Looking ahead, future releases may include Mythos-class models once safeguards prove effective, though no timeline is specified. Anthropic's emphasis on safety, utility and an ad-free model contrasts with competitors, positioning it to influence AI's role in secure, productive ecosystems.