Anthropic has officially launched Opus 4.7, a new flagship AI model designed to succeed Opus 4.6 with a specific focus on complex software engineering
Anthropic today announced its latest flagship model, Opus 4.7. As expected, Opus 4.7 brings significant improvements over Opus 4.6 in complex software engineering tasks. The startup claims that Opus 4.7 can handle long-running tasks more consistently while strictly following given instructions. Another notable upgrade over Opus 4.6 is its much-improved vision capability, allowing it to analyze higher-resolution images.
Anthropic highlighted that while Opus 4.7 performs better than Opus 4.6, it is still significantly less capable than Claude Mythos Preview. As shown in the benchmark comparison below, Opus 4.7 trails Mythos Preview by a wide margin in almost all benchmarks. Still, with Opus 4.7, Anthropic is once again leading agentic coding benchmarks, beating OpenAI’s GPT-5.4.
Unlike Mythos Preview, which is limited to a handful of organizations, Opus 4.7 is available today across all Claude products, the Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. Opus 4.7 is priced the same as Opus 4.6 at $15 per million input tokens and $75 per million output tokens.
Anthropic also said that Opus 4.7 includes built-in safeguards that automatically detect and block requests related to high-risk cybersecurity uses. Security professionals who want to use Opus 4.7 for legitimate cybersecurity purposes can apply to join Anthropic’s new Cyber Verification Program.
Anthropic noted that developers should plan for two changes when migrating from Opus 4.6 to Opus 4.7. First, Opus 4.7 uses an updated tokenizer that improves how the model processes text. This tokenizer uses roughly 1.0x to 1.35x more tokens depending on the content type. Second, Opus 4.7 tends to “think” more at higher effort levels, which can result in higher output token usage.
Alongside Opus 4.7, Anthropic today also announced a new /ultrareview slash command in Claude Code. It creates a dedicated review session that reads through changes and flags bugs and design issues. Claude Pro and Claude Max Code users will get three free ultrareviews during the preview phase.
