Introducing Claude Opus 5

Anthropic Releases Claude Opus 5, Prioritizing Cost-Effective Performance

Anthropic has announced the immediate availability of Claude Opus 5, its latest large language model. The new release is positioned as a highly practical and efficient model, delivering intelligence that approaches the company's frontier model, Claude Fable 5, at half the cost. Available now as the default on Claude Max and the top offering on Claude Pro, Opus 5 establishes a new state-of-the-art on key coding and knowledge work evaluations, including Frontier-Bench and GDPval-AA, though it still trails competitors like Mythos 5 on specialized cybersecurity tasks. The launch signals a clear focus on making high-end AI capabilities more accessible for daily development and enterprise workflows.

Performance and Cost-Effectiveness

Claude Opus 5 is engineered to provide a substantial performance uplift over its predecessor, Opus 4.8, without an increase in price. Customers can use a new 'effort setting' to balance intelligence against token consumption for faster, cheaper results. The model demonstrates significant gains on valuable software engineering and automation benchmarks.

• On Frontier-Bench v0.1, Opus 5 surpasses all other models and more than doubles the performance of Opus 4.8 at a lower cost per task.
• On CursorBench 3.2, it performs within 0.5% of Fable 5’s peak score but at half the cost.
• Its pass rate on Zapier AutomationBench, which measures end-to-end business task completion, is approximately 1.5 times higher than the next-best model for the same cost.
• On the OSWorld 2.0 computer use benchmark, it outperforms all other models at any given cost, surpassing Fable 5's best result at just over a third of the price.

A Focus on Agentic Work and Reliability

Beyond raw benchmarks, Anthropic emphasizes that Opus 5 exhibits greater 'agency' and 'thoroughness' by verifying its own work and iterating on complex problems. Early customer reports from firms like Zapier, Box, and Lovable highlight the model's consistency and its ability to handle long, multi-step tasks that previously required significant human oversight. For example, the model successfully wrote its own computer vision pipeline to solve a task when direct input was unavailable and built its own test harness to validate code for a new market data feed. This improved judgment and reliability are critical for developers and enterprises looking to deploy more autonomous agents for financial modeling, code generation, and business process automation.

The launch of Claude Opus 5 indicates a significant market shift from a pure race for frontier intelligence to delivering models with near-frontier capabilities that are economically viable and reliable enough for daily, mission-critical production workloads. The emphasis on cost-per-task, improved agentic judgment, and run-to-run consistency suggests the industry is maturing toward practical enterprise adoption and ROI.