ReleaseBytes
anthropic Anthropic News ·

Anthropic Releases Claude Opus 5 with Enhanced Performance and Efficiency

aigaengineer
feature patch

Anthropic has launched Claude Opus 5, a significant upgrade to its Opus tier model, offering improved performance in coding and professional tasks at a reduced cost compared to its predecessor. It sets new benchmarks on evaluations like Frontier-Bench and GDPval-AA, demonstrating enhanced efficiency for everyday use, particularly in software engineering and scientific research. Opus 5 is now the default model on Claude Max and the strongest on Claude Pro, benefiting users who require advanced capabilities for complex workloads.

  • →Claude Opus 5 offers state-of-the-art performance at lower cost
  • →Improved cost-effectiveness and performance tuning
  • →Excels in software engineering and problem-solving tasks
  • →Enhanced capabilities in scientific research and visual outputs
  • →Improved agency and thoroughness in complex tasks
Features (1) ›
  • Claude Opus 5 offers state-of-the-art performance at lower cost

    Claude Opus 5 is now available, delivering performance close to Claude Fable 5 but at half the price. It excels on coding and knowledge work evaluations, setting new state-of-the-art benchmarks while remaining behind Mythos 5 on cybersecurity tasks. The model is designed for daily use with improved efficiency, serving as the new default on Claude Max and the strongest option on Claude Pro.

Enhancements (5) ›
  • Improved cost-effectiveness and performance tuning

    Opus 5 provides significantly improved performance for the same cost as Opus 4.8. Users can adjust the model's 'effort' setting to optimize for intelligence or conserve tokens for faster, cheaper results. This allows for fine-tuning performance based on specific needs, offering flexibility for various workloads.

  • Excels in software engineering and problem-solving tasks

    The model demonstrates superior performance on valuable software engineering tasks, surpassing other models on Frontier-Bench v0.1 and showing comparable performance to Fable 5 on CursorBench 3.2 at a lower cost. It also shows strong results in knowledge work and problem-solving, with three times the score of the next-best model on ARC-AGI 3.

  • Enhanced capabilities in scientific research and visual outputs

    Opus 5 shows meaningful improvements for scientific research, particularly in life sciences and organic chemistry tasks like inferring molecular structures. It also demonstrates improved ability in producing stronger visual outputs compared to previous models.

  • Improved agency and thoroughness in complex tasks

    Claude Opus 5 is more adept at verifying its work and iterating to achieve success. Examples include autonomously developing a computer vision pipeline to reconstruct a machine part from raw pixels and identifying and fixing a subtle bug in an open-source package manager that competing models missed.

  • Significant gains across various professional workflows

    Early access customers report notable improvements in areas such as full-stack application development, debugging, data analysis, due diligence, and financial modeling. Opus 5 shows increased accuracy, efficiency, and better reasoning at lower effort levels across specialized enterprise content and legal agent work.

Read the original announcement →

https://www.anthropic.com/news/claude-opus-5

Related releases