Anthropic unveils new Claude Opus 4.5, its 'most intelligent' model (www.businessinsider.com)

Anthropic this week unveiled Claude Opus 4.5, which it bills as its most advanced and “most intelligent” model yet. Released just three months after the prior update, Opus 4.5 focuses on high-stakes workplace tasks — generating production-quality code, analyst-grade Excel spreadsheets, and polished presentations — and adds capabilities for more sophisticated, long-running AI agents. The release completes the Claude 4.5 family: Opus models aim for advanced reasoning and complex problem solving, while Sonnet and Haiku target speed and efficiency. The launch matters because it reinforces Anthropic’s push to lead enterprise AI: a recent Menlo Ventures report credits Anthropic with 32% enterprise adoption (vs. OpenAI 25%, Google 20%, Meta 9%). Anthropic says Opus 4.5 outperformed Google Gemini 3 Pro and OpenAI’s GPT-5.1 on the SWE-Bench Verified coding benchmark. Behind the scenes, Anthropic is locking in compute and capital — plans to spend $30B for Nvidia chips via Microsoft’s cloud (with Nvidia and Microsoft committing up to $10B and $5B respectively) — and is in talks with Google for new funding that could value the company at ~$350B. Those deals accelerate capacity for larger models but also raise questions about circular investments and market concentration as Anthropic scales.