Anthropic Launches Opus 5, Offering Near-Fable Performance at Lower Cost and Fewer Restrictions

On Friday, Anthropic launched Opus 5, the latest iteration of its long-standing flagship large language model. The company positions Opus 5 as a powerful and more accessible alternative to its top-performing Fable 5 model, offering comparable capabilities for a significantly lower price and with far fewer usage restrictions. The release comes just two months after Opus 4.8 debuted in late May, and rounds out the company’s “5 series” upgrades, leaving only the lightweight Haiku model awaiting an update.

Key takeaways

• Opus 5 offers performance that comes close to Anthropic’s most powerful commercial model, Fable 5, but at half the price.
• The new model is subject to significantly fewer safeguards, with safety classifiers expected to intervene around 85% less often than for Fable 5.
• Anthropic highlights Opus 5’s improved capabilities in complex coding, scientific research, and verifying its own work.
• Unlike Fable and Mythos 5, Opus 5 is not covered by Anthropic’s 30-day data retention policy, addressing privacy concerns.
• The company is introducing a beta “Automatic Fallbacks” feature to route blocked prompts to a less powerful model instead of returning an error.

Performance and Pricing

According to Anthropic, Opus 5 offers capabilities that nearly match those of Fable 5, the company’s most powerful commercially available model. The company claims Opus 5 outperforms Fable 5 on several benchmarks and is better at knowledge work and complex coding tasks than any of its predecessors, including Opus 4.8. It also excels in scientific research applications, such as predicting how protein sequence variations affect molecular function. Crucially, this performance comes at a substantially lower cost. API pricing for Opus 5 is set at $5 per million input tokens and $25 per million output tokens, which Anthropic states is half the price of accessing Fable 5.

Reduced Safeguards and New Features

A major differentiator for Opus 5 is its less restrictive safety framework. Anthropic says the model’s safeguards are “designed to allow beneficial uses” in areas like cybersecurity and biology. While some cyber-related guardrails have been strengthened—for instance, preventing the model from scanning software binaries for vulnerabilities—the overall system is much lighter. The company expects its safety classifiers to engage approximately 85% less often for Opus 5 than for Fable 5. To make safeguards less disruptive, Anthropic is rolling out a beta feature called Automatic Fallbacks. When enabled, this feature will automatically reroute a prompt that triggers a safety classifier to a less powerful model, providing a functional response instead of an error message. Furthermore, Opus 5 is explicitly excluded from the 30-day data retention policy that applies to Fable and Mythos models.

Why it matters

The launch of Opus 5 represents a strategic shift for Anthropic, aiming to provide a high-performance model that is both more affordable and less frustrating for developers than its top-tier offering. By addressing widespread user complaints about Fable’s heavy-handed restrictions and high cost, Anthropic is positioning Opus 5 as the go-to model for practical, everyday tasks. The introduction of features like Automatic Fallbacks and the exclusion from data retention policies directly respond to user feedback on usability and privacy. In a competitive landscape that now includes models like China’s Moonshot Kimi K3, which offers competitive pricing, Anthropic’s move to enhance the value proposition of its flagship Opus line is a critical effort to retain and grow its developer base.