Anthropic has launched Claude Opus 4.6, its most advanced artificial intelligence model to date. The release, announced on Thursday, introduces a new “Agent Teams” capability that allows multiple AI agents to collaborate on complex tasks. The launch has sparked significant movement in the technology sector, coinciding with a reported sell-off in financial and software stocks.
The new model represents a major shift in how businesses can utilize artificial intelligence, moving beyond simple question-and-answer interactions to autonomous, multi-step workflows. According to Anthropic, Opus 4.6 is designed to handle “deep work” that requires extended reasoning, planning, and execution.
Agent Teams Change the Workflow
The standout feature of Claude Opus 4.6 is “Agent Teams.” This system enables users to deploy a squad of specialized AI agents coordinated by a single team leader. Instead of a single model attempting to solve a problem in one go, the agents can divide the work, challenge each other’s assumptions, and debate potential solutions.
Early demonstrations indicate that this multi-agent approach produces higher-quality results than single-model executions. The agents are capable of autonomous collaboration, where teammates communicate with one another to refine their output. This structure allows the model to take on ambitious objectives, break them down into concrete steps, and execute them with minimal human oversight. For users, the experience is described as working with a capable collaborator rather than a passive tool.
Enhanced Reasoning and Benchmarks
Claude Opus 4.6 integrates “adaptive thinking,” a feature that allows the model to determine the necessary level of cognitive effort for a given task. This capability is supported by a new /effort parameter, which lets developers control the model’s reasoning intensity.
Anthropic reports that Opus 4.6 has achieved state-of-the-art scores on several key benchmarks:
- Humanity’s Last Exam: The model leads all other frontier models on this complex, multidisciplinary reasoning test.
- Terminal-Bench 2.0: It secured the highest score on this evaluation for agentic coding.
- GDPval-AA: The model demonstrated superior performance on economically valuable knowledge work in finance and legal domains.
To support these complex workflows, Anthropic has expanded the model’s context capabilities. Opus 4.6 introduces a “context compaction” feature, triggered at 50,000 tokens, which allows the model to process up to 3 million total tokens. This ensures that the AI retains critical information over long conversations and extensive projects without losing track of earlier details.
Market Reaction and Industry Impact
The release of Claude Opus 4.6 has had an immediate impact on the financial markets. Reports following the launch highlighted a broad sell-off in technology stocks, specifically targeting companies in the financial data and software sectors.
The market movement appears to be driven by investor sentiment regarding the capabilities of the new model. With Opus 4.6 performing exceptionally well on financial and legal benchmarks, there are growing questions about the future of traditional knowledge work software. Companies that provide financial data and enterprise software solutions saw their stock prices decline as the market adjusted to the potential disruption posed by these advanced autonomous agents.
Enterprise Availability
Claude Opus 4.6 is available immediately for enterprise use. Anthropic has partnered with major cloud providers to ensure broad access. The model has launched on Amazon Bedrock and is also available via Microsoft Foundry, allowing businesses to integrate these new capabilities directly into their existing workflows.
