Anthropic announced the release of its latest artificial intelligence model, Claude Opus 5, on July 24, 2026. The company stated that Opus 5 achieves capabilities close to its advanced Claude Fable 5 model but at half the price. This new model offers significant improvements in coding tasks and general knowledge work, positioning it as a more accessible option for complex AI applications.
Claude Opus 5 is now available across Anthropic's platforms and through partners like Amazon Web Services. Anthropic claims Opus 5 matches the intelligence of Claude Fable 5 in many areas, particularly in coding and knowledge work evaluations such as Frontier-Bench and GDPval-AA, where it sets a new standard. This advancement comes with the same pricing as its predecessor, Opus 4.8, at $5 per million input tokens and $25 per million output tokens. The company highlighted that Opus 5 is designed for efficiency, working faster and more effectively than previous models. It is now the default model for Claude Max users and the most powerful option for Claude Pro subscribers.
Anthropic's announcement detailed Opus 5's strengths in software engineering, describing it as a model that deeply understands codebases and can manage complex, multi-step tasks for feature development and bug fixing. On benchmarks like CursorBench 3.2, Opus 5 reportedly performs within 0.5% of Fable 5's peak performance but at half the cost. Similarly, on OSWorld 2.0, it surpasses Fable 5's results at just over a third of the expense. For knowledge-based tasks, Opus 5 shows a threefold increase in scores on evaluations like ARC-AGI 3 compared to the next best model. It also demonstrates a 1.5 times higher pass rate on Zapier AutomationBench for business task completion at a comparable cost.
While Opus 5 represents a significant leap, Anthropic noted that it remains behind their top-tier Mythos 5 model in cybersecurity tasks. However, its cyber capabilities are stated to exceed those of Opus 4.8. The safeguards implemented in Opus 5 are aligned with those of Claude Fable 5, with a notable change allowing source-code vulnerability discovery at all access levels. Anthropic also reported a substantial reduction in safety classifier triggers for Opus 5 compared to Fable 5, with a 5% flag rate on API calls versus Fable 5's 42%.
The release of Opus 5 is part of Anthropic's rapid iteration cycle, with this being the fourth Claude 5 model launched in under two months, following Mythos 5, Fable 5, and Sonnet 5. This cadence underscores a shift in the AI industry towards continuous improvement in capability, cost, and speed rather than solely focusing on large, singular product launches.
Opus 5 is available through the Claude API and on Amazon Bedrock, with features like zero data retention by default for enterprise users. The model also offers a "Fast mode" which runs approximately 2.5 times faster than the default speed, available at double the base price. Anthropic is also introducing beta features such as mid-conversation tool changes on the Claude Platform and automatic fallbacks on the API, which reroute flagged requests to other models to ensure continuous operation.
The company emphasized Opus 5's efficiency, noting that customers can optimize performance by adjusting its "effort setting," allowing for a balance between intelligence and token conservation for faster, cheaper results. This flexibility caters to a range of use cases, from demanding software engineering projects to everyday knowledge work. Developers can access Opus 5 via the Claude API and through major cloud providers including Amazon Web Services, Google Cloud, and Microsoft Foundry.