Anthropic Unveils Claude Opus 4.6: Revolutionary Agent Teams and Million-Token Context Transform Enterprise AI Landscape

News Summary

Anthropic unveiled Claude Opus 4.6 on Thursday, February 5, 2026 (PST), marking a significant advancement in enterprise AI capabilities. The new flagship model features a groundbreaking 1 million token context window, revolutionary "agent teams" functionality, and state-of-the-art performance across coding, financial analysis, and knowledge work benchmarks, positioning it as a direct competitor to OpenAI's GPT-5.2.

Anthropic Launches Claude Opus 4.6: Enterprise AI Takes Quantum Leap with Agent Teams and Million-Token Context

San Francisco, February 5, 2026 — Anthropic released Claude Opus 4.6 on Thursday morning (PST), introducing what the company describes as a paradigm shift in enterprise artificial intelligence. The latest iteration of its flagship model delivers unprecedented capabilities in autonomous task execution, extended reasoning, and collaborative AI workflows.

Revolutionary Context Window Expands AI Capabilities

Claude Opus 4.6 becomes the first model in Anthropic's Opus family to support a 1 million token context window, placing it alongside Google's Gemini models in the ultra-long context category. This massive expansion allows the model to process approximately 1,500 pages of text, 30,000 lines of code, or over an hour of video content in a single prompt.

The model demonstrated exceptional performance on the MRCR v2 benchmark, achieving 76% accuracy in needle-in-a-haystack information retrieval tasks, compared to just 18.5% for its predecessor, Claude Sonnet 4.5. According to Anthropic, this represents a qualitative shift in eliminating "context rot" — the degradation of model performance over extended conversations.

Agent Teams: Parallel Processing for Complex Workflows

The introduction of "agent teams" marks a fundamental architectural change in how Claude approaches complex tasks. Instead of sequential task execution by a single agent, Opus 4.6 can now deploy multiple specialized agents working in parallel, each handling distinct components while coordinating directly with one another.

Scott White, Anthropic's Head of Product for Enterprise, compared the functionality to managing a talented human team. "You can split the work across multiple agents — each owning its piece and coordinating directly with the others," White explained in an interview with TechCrunch. This capability is currently available in research preview for API users and subscription customers.

Benchmark Dominance Across Professional Domains

Claude Opus 4.6 established new performance records across multiple industry-standard evaluations:

Coding Excellence: The model scored 65.4% on Terminal-Bench 2.0, the highest score ever recorded on this agentic coding evaluation. It also leads competitors on OSWorld agentic computer use benchmark, scoring 72.7% compared to Opus 4.5's 66.3%.

Financial and Legal Analysis: On GDPval-AA, which measures performance on economically valuable knowledge work, Opus 4.6 achieved 1,606 Elo points — outperforming OpenAI's GPT-5.2 by approximately 144 Elo points and its predecessor by 190 points. The model also reached 90.2% on BigLaw Bench, the highest score of any Claude model on legal reasoning tasks.

Novel Problem-Solving: Perhaps most remarkably, Opus 4.6 scored 68.8% on the ARC AGI 2 benchmark, which tests problems that are easy for humans but notoriously difficult for AI systems. This represents an 83% improvement over Opus 4.5's 37.6% score.

Information Retrieval: The model achieved the highest industry score on BrowseComp, demonstrating superior ability to locate hard-to-find information across the web.

Microsoft Office Integration Deepens

Anthropic announced substantial upgrades to its Office suite integrations. Claude in Excel can now handle longer, more complex tasks and apply multi-step transformations in a single operation without requiring explicit structural explanations.

The company also unveiled Claude in PowerPoint as a research preview, enabling the AI to read existing slide layouts, fonts, and corporate templates, then generate or edit presentations that maintain brand consistency. This integration is available in beta to Max, Team, and Enterprise plan customers.

Advanced API Controls for Developer Flexibility

Opus 4.6 introduces several sophisticated features for API developers:

Adaptive Thinking: The model can autonomously determine when to employ deeper reasoning versus quick responses, using contextual clues to balance quality against latency and cost.

Effort Levels: Developers gain explicit control through four effort settings (low, medium, high, and max), allowing precise tradeoffs between intelligence, speed, and computational expense.

Context Compaction: A beta feature that automatically summarizes older conversation segments when context limits approach, enabling extremely long interactions without performance degradation.

Market Impact and Enterprise Adoption

The release triggered significant market reaction. Software stocks experienced substantial volatility earlier this week following Anthropic's announcement of industry-specific plugins for its Cowork tool. Thomson Reuters fell 15.83% on Tuesday, while LegalZoom dropped nearly 20%, as investors weighed the potential for AI to displace specialized research and financial analysis software.

Despite these market concerns, enterprise adoption continues to accelerate. According to a recent Andreessen Horowitz survey, 44% of enterprises now use Anthropic in production environments — the largest share increase of any frontier AI lab since May 2025.

Real-World Deployments Show Promise

Early access partners reported substantial productivity gains. Rakuten deployed Opus 4.6 to autonomously manage a 50-person organization, successfully closing 13 issues in a single day. Sarah Sachs, Head of AI at Notion, described the model as evolving beyond a tool to become "a truly capable collaborator."

Michael Truell, co-founder of AI coding platform Cursor, noted the model's persistence on challenging problems: "Claude Opus 4.6 excels on the hardest problems. It shows greater persistence, stronger code review, and the ability to stay on long tasks where other models tend to give up."

Pricing and Availability

Anthropic maintained its competitive pricing structure at $5 per million input tokens and $25 per million output tokens. The model is immediately available through claude.ai, the Claude API (model ID: claude-opus-4-6), and all major cloud platforms including Amazon Web Services, Google Cloud, and Microsoft Azure.

The model is also being integrated into GitHub Copilot and is gradually rolling out to Copilot Pro, Pro+, Business, and Enterprise users.

Safety and Alignment Commitments

According to Anthropic's extensive system card, Opus 4.6 maintains an overall safety profile equal to or better than any other frontier model, with low rates of misaligned behavior across safety evaluations. The company emphasizes that safety has not been sacrificed for performance gains.

Industry Context and Competition

The launch comes just 72 hours after OpenAI's Codex release, underscoring the accelerating pace of competition in AI development tools. White told media outlets that Anthropic has transitioned Claude from "a model that you can sort of talk to accomplish a very small task" to "something that you can actually hand real significant work to."

The release positions Anthropic for what White termed the "vibe working" era, where knowledge workers increasingly delegate substantive professional tasks to AI systems capable of autonomous execution with minimal supervision.

Claude Opus 4.6 represents Anthropic's most ambitious enterprise AI offering to date, combining technical advances in context handling, parallel agent coordination, and domain-specific expertise to challenge the prevailing assumptions about AI's role in professional workflows.