Anthropic ships Opus 4.7

Anthropic released Claude Opus 4.7 on April 16, making its newest Opus model generally available across Claude apps, the API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. (anthropic.com) Large language models predict the next chunk of text, and newer versions are being tuned to stay on task longer, use tools, and work from images as well as text. Anthropic said Opus 4.7 improves advanced software engineering, long-running coding work, instruction-following, and image understanding over Opus 4.6. (anthropic.com) (support.claude.com) Anthropic also launched “task budgets” in public beta for developers building autonomous agents, which are programs that can plan and take multiple steps on their own. The company described the feature as a way to set spending ceilings for agent runs as customers push models into longer workflows. (venturebeat.com) (siliconangle.com) The release lands nine days after Anthropic published its Claude Mythos Preview system card on April 7. In that document, Anthropic called Mythos Preview its most capable frontier model and said it would not make it generally available, limiting it instead to a defensive cybersecurity program with selected partners. (anthropic.com) Anthropic tied Opus 4.7 directly to that decision. In its launch post, the company said Opus 4.7 is “less broadly capable” than Mythos Preview and is the first model on which it is testing new cyber safeguards before any broader release of Mythos-class systems. (anthropic.com) Those safeguards reflect what Anthropic disclosed about Mythos Preview’s cyber performance. In a technical post on April 7, the company said Mythos Preview could identify and exploit zero-day vulnerabilities in major operating systems and web browsers during testing, and said more than 99% of the vulnerabilities it found had not yet been patched. (red.anthropic.com) For customers, Opus 4.7 is the model Anthropic is willing to sell broadly now. Anthropic said pricing stays at $5 per million input tokens and $25 per million output tokens, the same rates as Opus 4.6, and developers can call it with the model name `claude-opus-4-7`. (anthropic.com) Anthropic’s public framing is that Opus 4.7 gives developers more capability in coding and vision while keeping tighter controls around the stronger Mythos line. The next test is whether those new safeguards and budget controls hold up once customers put the model into real production agent workflows. (anthropic.com) (venturebeat.com)