← All models

Claude Opus 4.7

Published Publisher Anthropicclaude-opus-4-7

Claude Opus 4.7

Positioning and release

Released on April 16, 2026, Claude Opus 4.7 is an Anthropic model for complex reasoning, advanced software engineering, and long-running agentic work. At launch, Anthropic described it as an improvement over Opus 4.6, while noting that it was less broadly capable than Claude Mythos Preview. (anthropic.com)

As of November 16, 2026, Anthropic lists it as active, legacy, with retirement no earlier than April 16, 2027. (platform.claude.com)

Capabilities and uses

  • Accepts text and images and produces text. It has a 1 million-token context window and a standard maximum output of 128,000 tokens; the Batch API beta supports up to 300,000 output tokens. It supports adaptive thinking, with high as the default effort level. Its knowledge cutoff is January 2026. (platform.claude.com)
  • Designed for complex coding, debugging, research agents, document analysis, and professional knowledge work. Anthropic reports improvements in instruction following, long-running task consistency, and verification of outputs compared with Opus 4.6. (anthropic.com)
  • Supports higher-resolution image input, with images up to 2,576 pixels on the long edge. Example uses include reading dense screenshots, extracting information from diagrams, and computer-use workflows. (anthropic.com)
  • Available through the Claude API and listed on Amazon Bedrock, Google Cloud, and Microsoft Foundry; availability and billing conditions depend on the platform. (platform.claude.com)

Limitations and safety

  • Outputs can be incorrect or incomplete and should be checked, particularly for consequential decisions or autonomous actions. Anthropic's system card documents safety and alignment risks, including rare observed attempts to circumvent restrictions. (www-cdn.anthropic.com)
  • Safeguards detect and block prohibited or high-risk cybersecurity requests; legitimate security work may be subject to verification. (anthropic.com)
  • Compared with Opus 4.6, the updated tokenizer can turn the same input into approximately 1.0–1.35 times as many tokens, depending on content. Higher effort can also increase output-token use, so unchanged per-token prices do not guarantee unchanged task costs. (anthropic.com)

Claude API pricing

The following Anthropic list prices are in USD per 1 million tokens, apply across the model's context window, and were checked on November 16, 2026. Cache-write and cache-read prices apply to the relevant input tokens; output tokens are billed separately. These are Claude API rates, not a claim that other platforms have identical terms. (assets.claude.com)

Service and inference scope Base input Output 5-minute cache write 1-hour cache write Cache read/refresh
Standard, global $5.00 $25.00 $6.25 $10.00 $0.50
Standard, US-only inference $5.50 $27.50 $6.875 $11.00 $0.55
Batch, global $2.50 $12.50 $3.125 $5.00 $0.25
Batch, US-only inference $2.75 $13.75 $3.4375 $5.50 $0.275

Official sources