Claude Opus 4.5 arrives with new controls and lower pricing

Anthropic released its newest flagship model, Claude Opus 4.5. The company stated the model sets a new standard for complex software development, automated task execution, and computer usage. The release comes with new controls for developers and a more accessible price point for advanced capabilities.

Major gains in coding and automation

Opus 4.5 focuses heavily on professional software engineering and automated workflows. Anthropic reports the model achieves superior performance on industry coding benchmarks. It scored 80.9% on the verified SWE-bench test, showing a high capacity for solving engineering problems.

The model now handles autonomous, long-horizon coding projects. Developers find that multi-day assignments can now take only a few hours. The system generates better code, covers more test cases, and makes cleaner architectural decisions during code cleanup.

For automated tasks, known in the field as agentic work, the model shows a marked improvement in decision-making and planning. It executes complex, multi-step workflows with fewer errors than its predecessors. The computer use feature includes a new “zoom” action, allowing the model to inspect specific screen regions at full resolution for greater detail when navigating a desktop environment.

New controls and cost structure

Anthropic introduced the Effort Parameter to give developers greater control over the model’s performance and cost. Opus 4.5 is the only model that supports this feature, allowing users to trade thoroughness for token usage. Developers can set the parameter to:

• High Effort: Maximum detail for deep analysis.

• Medium Effort: A balanced approach for most general tasks.

• Low Effort: Most token-saving response for high-volume automation.

The company also lowered the price for its most capable model. Opus 4.5 costs USD 5 per million input tokens and USD 25 per million output tokens. This makes the powerful Opus-class intelligence available at about one-third the cost of earlier versions. The company also improved its context management, summarizing earlier messages to allow for extended, long-running conversations without hitting length limits.

Claude Opus 4.5 is immediately available to Pro, Max, Team, and Enterprise subscribers. Developers can access the model through the Claude Developer Platform API and via major cloud platforms including Amazon Bedrock, Google Cloud’s Vertex AI, and Microsoft Foundry.