Scale News

Anthropic’s Opus 5 nears Fable 5 in scope

 ·  By Isadora Dunmore
Anthropic’s Opus 5 nears Fable 5 in scope - ai model
Anthropic’s Opus 5 nears Fable 5 in scope

Anthropic launched Opus 5, a new AI model that matches the performance of its flagship Fable 5 at half the cost while enhancing coding, knowledge work, and autonomous workflows.

Performance at a lower cost

Opus 5 serves as the default model for Claude Max subscribers and the top-tier option for Claude Pro users. It eliminates the 30-day data retention requirement of Fable 5, making it more suitable for regular use. Pricing remains the same as Opus 4.8—$5 per million input tokens and $25 per million output tokens.

The model surpasses Fable 5 on most benchmarks, including business workflows and agentic tasks. On the GDPval-AA v2 benchmark, it achieved a score of 1861, outperforming Fable 5 (1747) and OpenAI’s GPT-5.6 Sol (1736). It also leads in Zapier’s AutomationBench, where its pass rate doubles that of the next-best model at the same cost per task.

While Fable 5 maintains an advantage in long-term autonomy and specialized tasks like legal benchmarks, Opus 5 offers a more practical solution for daily work. Sonnet 5, introduced weeks earlier as the most efficient for everyday tasks, now occupies an awkward middle ground between the cheaper Haiku 4.5 and the more advanced Opus and Fable models.

Safety and practical trade-offs

Opus 5 incorporates stricter safety filters than its predecessor. Anthropic avoided training it on cybersecurity tasks, though it still identifies vulnerabilities in open-source code without the ability to exploit them.

Related: Things To Consider When Redesigning Your Website

API users can enable fallback options, while enterprises and researchers in Anthropic’s Cyber Verification Program access a less restricted version.

The release aligns with an industry-wide focus on cost efficiency. Major AI labs now prioritize price-performance, and Opus 5 addresses this demand. OpenAI’s GPT-5.6 Sol, for instance, costs $5/$30 per million tokens—slightly more than Opus 5—and only exceeds its performance on DeepSWE v1.1, an agentic coding test.

Cursor, an AI coding assistant, reported that Opus 5 delivers intelligence close to Fable 5 at Opus speed and cost, with only a 0.5% gap in peak performance for half the price.

Opus 5 is now available on all paid Claude plans and the Claude Platform as claude-opus-5. Anthropic anticipates it will become the preferred choice for enterprise users, though Fable 5 remains better suited for high-stakes, long-running projects.

Leave a Comment

Your email address will not be published.