source: Simon Willison: Introducing Claude Opus 5
level: technical
Anthropic released Claude Opus 5, a new model described as thoughtful and proactive, with intelligence close to the frontier Claude Fable 5 but at half the price. It leads the Artificial Analysis leaderboard, even ahead of Fable 5. The model is priced the same as Opus 4.8 and offers a fast mode at double the base cost. Early anecdotes suggest it can independently solve problems, such as building a computer vision pipeline when denied direct access to an image.
The model shows improved cybersecurity vulnerability detection, approaching Mythos 5 in finding flaws, but it remains substantially behind in exploiting them. Anthropic intentionally avoided training Opus 5 on cyber exploitation tasks, aiming to limit material threats. This design choice may address regulatory concerns. The release includes a new prompting guide and a companion piece on context engineering for Claude 5 generation models, signaling a shift in how users interact with the model.
Claude Opus 5 builds on the Opus line, balancing capability and cost. Its proactive behavior, demonstrated by autonomously reconstructing a 3D model from raw pixels, hints at more agentic AI. The model's performance and pricing could pressure competitors. For developers, the fast mode offers a speed trade-off, while the prompting guide suggests new interaction patterns. The release continues Anthropic's focus on safety, deliberately limiting cyber exploitation skills despite general capability gains.
why it matters: Claude Opus 5 offers near-frontier AI performance at half the cost, making advanced capabilities more accessible for developers and businesses while maintaining safety constraints on cyber exploitation.