Anthropic released Claude Opus 4.7, the most capable Opus model yet, with stronger instruction-following, self-verification of outputs, and better handling of long-running tasks; the release frames it as ideal for software engineering, long-horizon analysis, and agentic workflows.

Opus 4.7 launched on April 16, 2026, with improved benchmarks and broader availability across enterprise channels.

Its performance is positioned as the pinnacle of Claude models to date, emphasizing reliability, memory for multi-session work, and enhanced reasoning for complex tasks.

Early feedback notes the model is more literal in following instructions, which may require re-tuning prompts, and tokenizer updates can increase input token usage by roughly 1.35x.

Availability remains through Claude API, Amazon Bedrock, Google Vertex AI, and Microsoft Foundry, with a 1 million token context window.

Reasoning and knowledge tasks are strengthened, including handling ambiguous instructions, structured outputs, higher-quality documents and presentations, and long-context retention up to one million tokens.

Vision capabilities now process high-resolution images up to 2,576 pixels on the long edge, with higher token consumption for such inputs, enabling detailed analysis of diagrams, UI screenshots, and documents.

Safety enhancements include automatic safeguards against high-risk cybersecurity misuse, a Cyber Verification Program for professionals, and reductions in hallucinations and improved honesty.

Pricing remains at $5 per million input tokens and $25 per million output tokens, with a public beta task budget feature to manage long autonomous runs.

User testimonials highlight faster development velocity, better async workflow handling, more reliable long-running tasks, improved documentation, and strong enterprise benchmark performance.

Multimodal upgrades provide substantially better vision and extended capabilities for complex interfaces, diagrams, and life-sciences workflows.

New features include a higher effort level for hard problems, task budgets on Claude Platform, an ultrareview command in Claude Code for code reviews, and extended Auto mode for longer-running tasks.