Anthropic Unleashes Claude Opus 5: Near-Fable Intelligence at Half the Price, Reshaping the AI Landscape

Anthropic's Strategic Shift: Performance Meets Affordability

Anthropic has officially released Claude Opus 5, a powerful new AI model that aims to redefine the balance between advanced capabilities and cost-effectiveness in the rapidly evolving artificial intelligence landscape. The launch of Opus 5 marks a significant pivot for Anthropic, moving beyond raw capability leaps to focus on the economics of daily AI use. This latest iteration delivers intelligence that closely rivals Anthropic's top-tier Claude Fable 5 model, but at a substantially reduced price point. The introduction of Opus 5 follows a rapid succession of model releases from Anthropic, with four Claude 5 models launched in less than two months, highlighting the industry's accelerated pace of innovation. This strategic move positions Opus 5 as a compelling option for developers and enterprises seeking high-performance AI without the premium cost typically associated with frontier models.

Unpacking Opus 5's Enhanced Capabilities and Benchmarks

Claude Opus 5 is engineered for demanding applications, showcasing considerable improvements in several key areas. It is particularly strong in agentic coding, demonstrating an ability to understand and navigate complex codebases, sustain long-running tasks, and produce production-quality code with minimal oversight. On benchmarks such as Frontier-Bench v0.1, Opus 5 has surpassed all other models, including Fable 5, and more than doubled the performance of its predecessor, Opus 4.8, at a lower cost per task. Similarly, on CursorBench 3.2, it performs within 0.5% of Fable 5's peak score while costing half as much per task. Beyond coding, Opus 5 excels in knowledge work and problem-solving tasks. On ARC-AGI 3, an evaluation designed to test novel problem-solving, Opus 5's score is three times higher than the next-best model. It also demonstrates superior performance on Zapier AutomationBench, which measures a model's ability to complete business tasks from start to finish, achieving a pass rate approximately 1.5 times higher than the next-best model for the same cost. Furthermore, Opus 5 has shown significant gains in scientific research, particularly in biology tasks like inferring molecular structures from spectroscopy data and predicting protein function.

Key Features and Technical Advancements

Opus 5 comes with a 1M token context window as both its default and maximum, and features 128k max output tokens. A notable enhancement is the "thinking on by default" setting, which allows the model to convert additional effort into better results more reliably. Other new features include:

• Mid-conversation tool changes (beta): Allows adding or removing tools during a conversation while preserving the prompt cache.
• Default fallbacks mode (beta): Automatically applies Anthropic's recommended fallback models by refusal category.
• Lower prompt cache minimum: The minimum cacheable prompt length is now 512 tokens, down from 1,024 tokens on Opus 4.8.
• Fast mode (research preview): Available on the Claude API, offering 2.5 times the default speed at twice the base price.

Pricing, Availability, and Enterprise Integration

Claude Opus 5 is priced at $5 per million input tokens and $25 per million output tokens, maintaining the same cost as its predecessor, Opus 4.8. This pricing, combined with its near-Fable 5 performance, makes it a highly attractive option for a wide range of users. The model is widely available across various platforms. Developers can access it via the Claude API. It is also integrated into major cloud platforms, including Amazon Bedrock, where it benefits from zero data retention by default, and Microsoft Foundry, offering enhanced security and governance for enterprise users. Furthermore, Claude Opus 5 is now available in GitHub Copilot, designed for complex, long-running coding tasks and agentic workflows. For enterprise workflows, Opus 5 is built to handle complex, multi-day projects with deep reasoning and high accuracy, particularly for document-heavy tasks in sectors like financial services. It can power dependable, long-running agents that work for hours, navigating obstacles and recovering from errors to achieve their objectives.

Safety and Alignment

Anthropic has implemented robust safeguards for Opus 5, similar to those applied to Opus 4.8, with stronger guardrails on certain cyber tasks. The model's cyber classifiers are less restrictive than those on Fable 5, allowing it to find vulnerabilities in source code while blocking activities like binary-based vulnerability scanning and exploit generation. Opus 5 has also shown improvements in agentic safety, particularly in prompt injection robustness across coding, computer use, and browser use. Anthropic states that Claude Opus 5 is its most aligned model to date on its automated behavioral audit.