Claude Opus 4.7 launches

Anthropic released Claude Opus 4.7 on April 16, saying the model is now generally available and tuned for harder coding and longer autonomous tasks. (anthropic.com) Large language models predict the next token in a sequence, but newer versions are being sold as software that can keep working through a chain of steps with less supervision. Anthropic said Opus 4.7 improves on Opus 4.6 in advanced software engineering, long-horizon agentic work, knowledge work, vision tasks, and memory tasks. (anthropic.com) (platform.claude.com) Anthropic also changed the way developers can use the model. The company said Claude Code can now run in the background for long-running coding jobs, and the API model name is `claude-opus-4-7`. (anthropic.com) (platform.claude.com) The pricing stayed in Anthropic’s premium tier: $5 per million input tokens and $25 per million output tokens, with prompt caching discounts of up to 90% and batch-processing discounts of 50%. Anthropic said Opus 4.7 is available in Claude for Pro, Max, Team, and Enterprise, and through the Claude Platform, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. (anthropic.com) Anthropic is pushing the model toward office and engineering work that unfolds over hours, not single prompts answered in seconds. Its product page says Opus 4.7 is built for professional software engineering, complex agentic workflows, and high-stakes enterprise tasks, with “adaptive thinking” that uses more computation on harder problems. (anthropic.com) That pitch fits a broader race among model vendors to sell reliability, not just chatbot fluency. When Anthropic launched Claude 4 in May 2025, it described Opus 4 as a model for sustained performance on complex, long-running tasks and agent workflows; Opus 4.7 extends that same framing with a narrower upgrade cycle. (anthropic.com 1) (anthropic.com 2) Anthropic’s own documentation says teams may need to adjust prompts and test harnesses after the upgrade. The company said instruction following is stronger, which can improve compliance with detailed directions but also make the model behave more literally than earlier versions. (platform.claude.com) Vision is another part of the release. On April 17, one day after the model launch, Anthropic introduced Claude Design in research preview and said the product runs on Claude Opus 4.7, calling it the company’s most capable vision model. (anthropic.com) Anthropic is presenting Opus 4.7 as a model that can be left alone longer without drifting off task. The next test is whether developers and enterprise buyers treat that claim as strong enough to hand over more real work. (anthropic.com)