Anthropic has launched its latest AI model, Claude Opus 5, setting a new benchmark in intelligence and affordability. Claude Opus 5 is positioned as a more cost-effective alternative to the Claude Fable 5 model, achieving remarkable performance in coding and knowledge work tasks, including leadership in competitions like Frontier-Bench and GDPval-AA. However, Mythos 5 still holds a slight edge over Opus 5 in cybersecurity functions. This new model is expected to deliver improvements for everyday use on the Claude Max and Pro platforms.
The release marks a significant enhancement over its predecessor, Opus 4.8, bringing superior efficiency at a comparable cost. Opus 5 doubles the performance of Opus 4.8 on Frontier-Bench while cutting costs. On CursorBench, Claude Opus 5 performs nearly on par with Fable 5 but at just half the expense, leading on several frontiers such as software engineering, automation, and computer use benchmarks. Its prowess also extends to knowledge tasks, surpassing competitors by substantial margins across various evaluations.
Claude Opus 5’s proficiency in scientific research is noteworthy, especially in life sciences disciplines like organic chemistry and bioinformatics. It scores significantly higher than Opus 4.8 on tasks such as spectroscopy and protein function prediction.
The model excels in creative problem-solving, demonstrating agency and thoroughness in tasks without direct visual data, like drafting 3D models from pixelated drawings and identifying complex software bugs overlooked by others. Users report successful scenarios ranging from finance to legal domains, where Opus 5 outmatches previous models’ performance and precision.
Security and alignment have been priorities in this development. Opus 5 underwent strict evaluations to ensure safe operations, retaining safeguards from previous iterations against dual-use risks in scientific research and cybersecurity. It remains behind Mythos 5 for developing cybersecurity exploits, though it can effectively identify vulnerabilities.
Claude Opus 5 is priced similarly to Opus 4.8 but is available in Fast mode at increased speed for twice the cost. Anthropic also introduces beta features alongside the launch: modifiable mid-conversation tools and automatic model fallbacks for flagged requests.
As part of a clear commitment to pushing AI frontiers while ensuring secure applications, Anthropic’s Opus 5 opens new opportunities for developers and businesses looking to harness robust AI capabilities efficiently.
You can read the original article here: [https://www.anthropic.com/news/claude-opus-5](https://www.anthropic.com/news/claude-opus-5)
