Anthropic has launched Claude Opus 5, its latest enterprise AI model, positioning it as a lower-cost alternative to its flagship system while topping coding and business-automation benchmarks.
The model went live on 24 July 2026, priced at US$5 per million input tokens and US$25 per million output tokens — unchanged from its predecessor, Opus 4.8. Anthropic said Opus 5 is now the default model on Claude Max and the strongest available on Claude Pro.
Claude Opus 5 tops coding and automation benchmarks
On Frontier-Bench v0.1, an evaluation of software engineering ability, Opus 5 more than doubled the score of Opus 4.8 at a lower cost per task, the company said. On CursorBench 3.2, the model came within 0.5 per cent of its flagship Fable 5 model’s peak score at maximum effort, while costing half as much per task.
Scott Wu, chief executive of Devin developer Cognition, said the model:
approaches Fable-level performance at half the cost
The model also topped the AutomationBench leaderboard run by workflow automation firm Zapier, completing a full churn-prevention sequence — from flagging at-risk accounts to alerting account owners — without using more tokens than earlier Claude models, according to the company’s chief executive, Wade Foster.
Other early-access partners, including Cursor, Lovable and enterprise content platform Box, reported similar gains in debugging accuracy, build consistency and data analysis workflows.
Alignment gains, but cybersecurity limits remain
Anthropic said its automated behavioural audit found Opus 5 to be its most aligned model to date, adhering more closely to the company’s constitution than Opus 4.8, Sonnet 5 or Fable 5, with the lowest rate of deceptive behaviour among its recent releases.
However, the company said Opus 5 remains behind Mythos 5, its most capable model, on offensive cybersecurity and long-running biology research tasks. While Opus 5 approaches Mythos 5’s ability to identify software vulnerabilities, it lags well behind on developing exploits from those vulnerabilities, according to results from Anthropic’s OSS-Fuzz evaluation.
- Cyber classifiers on Opus 5 are set to intervene around 85 per cent less often than on Fable 5
- The model can find vulnerabilities in source code but is blocked from binary-based scanning, penetration testing and exploit generation
- Flagged requests fall back automatically to Opus 4.8 by default across Claude.ai, Claude Code and Claude Cowork
Availability
Claude Opus 5 is available today across Anthropic’s platforms, including the API, and is also offered in a Fast mode running roughly 2.5 times its default speed at twice the base price. Alongside the launch, Anthropic introduced mid-conversation tool changes and automatic model fallbacks in beta on its developer platform.
