Anthropic releases Claude Opus 5 with performance near Fable 5 at half the cost

Anthropic has launched Claude Opus 5, a new flagship AI model designed to provide performance levels approaching its top-tier Claude Fable 5 model. The company positions the release as a more cost-effective alternative for enterprises, pricing the model at $5 per million tokens for input and $25 per million tokens for output. This represents a 50 percent reduction in cost compared to the Fable 5 model, which is priced at $10 and $50 per million tokens for input and output, respectively.

According to benchmarks provided by Anthropic, Opus 5 demonstrates significant improvements over its predecessor, Opus 4.8, particularly in agentic coding, computer use, and long-horizon knowledge work. While Fable 5 maintains a slight edge in highly specialized cybersecurity and biology tasks, Opus 5 is reported to be more capable across a broader range of general business workflows. The model is also described as Anthropic's most aligned release to date, with a reduced likelihood of triggering safety guardrails.

To optimize the model for production environments, Anthropic has streamlined the system prompt, removing 80 percent of the previous Claude Code guidance. Engineers can utilize a new 'claude doctor' command to assist in updating existing prompts and skills to match the new architecture. Anthropic notes that users may experience more verbose responses from the model by default and suggests that those requiring shorter outputs should explicitly define length constraints in their prompts.

The company is also introducing an automated fallback function for its API. This feature is designed to route requests that are deemed too sensitive or dangerous for Opus 5 or Fable 5 to a less restrictive model, rather than blocking the interaction entirely. Additionally, Anthropic has emphasized that the new model does not require data retention, a feature intended to appeal to business customers concerned with privacy and compliance.

Independent analysis from Artificial Intelligence Index suggests that the cost-to-performance ratio for Opus 5 is competitive within the current market. The weighted average cost per task is estimated at $2.03 for Opus 5, compared to $2.75 for Fable 5 and $1.04 for OpenAI's GPT-5.6 Sol. The model is available immediately across all platforms and through Amazon Bedrock.