Anthropic launches Claude Opus 5

San Francisco-based AI research lab Anthropic has officially unveiled Claude Opus 5, a high-performance artificial intelligence model engineered to bridge the gap between frontier-level capability and everyday operational costs. Positioned directly beneath the company's flagship Fable 5, the new release aims to deliver near-top-tier intelligence at approximately half the price per task. As enterprise adoption matures and corporate leaders demand clear financial returns on AI infrastructure, Anthropic's latest move reflects a broader strategic realignment across the technology sector: moving away from raw performance records toward cost-effective, scalable deployment.

At the heart of the Claude Opus 5 launch is a calculated focus on computational efficiency. The model is priced at five dollars per million input tokens and twenty-five dollars per million output tokens, matching the baseline rates of its predecessor, Opus 4.8, while yielding substantially higher capability across multiple domain benchmarks. Crucially, Anthropic claims that Opus 5 achieves up to double the performance of Opus 4.8 on specialized tasks like software engineering while consuming significantly fewer tokens. Furthermore, users can activate a specialized Fast Mode, enabling the model to operate at two point five times its default speed for time-sensitive enterprise applications.

Benchmark results released alongside the announcement highlight remarkable gains in practical problem-solving and autonomous agency. On software engineering evaluations such as Frontier-Bench v0.1 and CursorBench 3.2, Opus 5 matched or closely approached the maximum reasoning scores of Fable 5 while maintaining a vastly smaller financial footprint. The model also demonstrated dramatic improvements on non-traditional logic and business automation tasks. On the ARC-AGI 3 benchmark, which measures novel reasoning capabilities, Opus 5 scored three times higher than competing models. Similarly, on Zapier AutomationBench, it completed multi-step workflow automation tasks with a pass rate nearly one point five times higher than alternatives at equivalent cost.

Early testers and corporate enterprise partners have documented several real-world scenarios where Opus 5 demonstrated unprecedented problem-solving autonomy. In one internal benchmark testing visual hardware modeling, the system was tasked with converting a technical drawing into a three-dimensional computer-aided design model without direct image access. Rather than failing, Opus 5 built its own computer vision analysis pipeline from scratch to process raw pixels and successfully recreated the component. In another case, the model pinpointed a subtle edge-case flaw in an open-source package manager that human maintainers and rival AI tools had overlooked, writing a comprehensive patch rather than a superficial fix.

The introduction of Claude Opus 5 comes at a pivotal moment for both Anthropic and the broader artificial intelligence industry. Enterprise buyers are increasingly scrutinized by financial leadership to demonstrate measurable productivity gains rather than indulging in experimental technology deployment. By optimizing Opus 5 for high-volume daily use, Anthropic is directly targeting routine business operations, coding pipelines, and automated customer support systems. The model now serves as the primary default driver for Anthropic's flagship enterprise subscriptions, including Claude Max and Claude Pro, providing organizations with an immediate path to lower API expenditures without sacrificing output quality.

Market analysts note that the release intensifies an already fierce competitive landscape. As rivals like OpenAI, Google, and emerging open-weights laboratories in China refine their own cost-effective models, Anthropic's dual focus on alignment, token efficiency, and safety positions Opus 5 as a strong contender for corporate workloads. Furthermore, with Anthropic reportedly preparing for a high-profile initial public offering, the debut of Opus 5 sends a clear signal to investors: commercial viability, predictable unit economics, and practical enterprise utility are now the primary benchmarks defining the next phase of the artificial intelligence revolution.