Anthropic released Claude Opus 5 on July 24, 2026. In its announcement, Anthropic says that “on coding and knowledge work evaluations like Frontier-Bench and GDPval-AA, Opus 5 is the new state-of-the-art, though it remains behind Mythos 5 on cybersecurity tasks.”

On this page

Availability and model ID

Anthropic says “Claude Opus 5 is available today on all platforms.” Developers call it as claude-opus-5 on the Claude API, and Anthropic points to Claude on AWS and Google Cloud as cloud options. In Anthropic’s own products, Opus 5 is available in Claude.ai, Claude Code and Claude Cowork. Anthropic says it is “the new default model on Claude Max, and the strongest model on Claude Pro.”

Anthropic describes Opus 5 as “designed to be used every day,” adding that “it works more efficiently than other models.”

New features

Anthropic lists several features shipping with Opus 5:

  • changing tools mid-conversation (beta)
  • automatic fallbacks for requests flagged by safety systems (beta)
  • a fast mode running at about 2.5 times the default speed
  • effort settings to trade performance against token use

Anthropic’s performance claims

Anthropic published its comparisons mostly as ratios rather than raw scores. All of the following are Anthropic’s own claims:

  • Frontier-Bench v0.1: “Opus 5 surpasses all other models, and more than doubles Opus 4.8’s performance.”
  • ARC-AGI 3: “Opus 5’s score is three times as high as the next-best model.”
  • Zapier AutomationBench: “Opus 5’s pass rate is around 1.5× the next-best model.”

Lab-published results for Opus 5 and later models are tracked on our benchmark leaderboard.

Behind Mythos 5 on exploits

Anthropic draws a specific line on cybersecurity: “Although Mythos 5 and Opus 5 identify vulnerabilities with similar success, Opus 5’s score on the development of exploits is far behind that of Mythos 5.” By Anthropic’s measure, Opus 5 matches Mythos 5 on finding vulnerabilities but not on turning them into working exploits.

On alignment, Anthropic reports that Opus 5 scores lowest among its recent models on its measures of misaligned behavior.

What it changes

Opus 5 became the default model on Claude Max at launch. For API developers, the beta features for mid-conversation tool changes and automatic safety fallbacks affect how agent applications are built on the model, and the effort settings let them trade performance against token use. Anthropic followed Opus 5 on September 22, 2026 with Claude Opus 5.5.

Opus 5 is listed on our AI model release timeline.