Anthropic Ships Claude Opus 4.6, Tightening the Race With OpenAI

When Anthropic’s engineers sat down to test their latest model on real-world coding tasks, something unexpected happened. Claude Opus 4.6 didn’t just complete the assignments—it anticipated problems the developers hadn’t even noticed yet. In one internal test, the model autonomously closed 13 issues and assigned 12 more to the right team members across six different repositories, managing a 50-person organization in a single day.

“Claude Opus 4.6 is the strongest model Anthropic has shipped. It takes complicated requests and actually follows through, breaking them into concrete steps, executing, and producing polished work even when the task is ambitious.” — Notion Engineering Team

A Million Tokens of Context

Anthropic announced Claude Opus 4.6 on Tuesday, marking a significant escalation in the AI capabilities race. For the first time in an Opus-class model, Anthropic is offering a 1 million token context window in beta—enough to process roughly 750,000 words in a single pass. That’s the equivalent of analyzing the entire Lord of the Rings trilogy alongside detailed technical documentation and still having room for conversation.

The coding improvements are substantial. Opus 4.6 plans more carefully, sustains agentic tasks for longer periods, and operates more reliably in larger codebases. The model demonstrates better code review and debugging skills, often catching its own mistakes before they become problems. On Terminal-Bench 2.0, an evaluation of agentic coding capabilities, Opus 4.6 achieved the highest score of any model tested.

Knowledge work performance shows equally impressive gains. On GDPval-AA—an evaluation measuring performance on economically valuable tasks in finance, legal, and other professional domains—Opus 4.6 outperformed OpenAI’s GPT-5.2 by approximately 144 Elo points and its own predecessor by 190 points. For legal reasoning specifically, the model achieved a 90.2% score on BigLaw Bench, with 40% perfect scores and 84% above 0.8.

“Claude Opus 4.6 reasons through complex problems at a level we haven’t seen before. It considers edge cases that other models miss and consistently lands on more elegant, well-considered solutions.” — Cognition Labs

Agent Teams and Autonomous Workflows

The release introduces several features designed for enterprise deployment. In Claude Code, users can now assemble agent teams to work on tasks collaboratively. On the API, Claude can use compaction to summarize its own context, enabling longer-running tasks without hitting limits.

Adaptive thinking allows the model to pick up on contextual clues about how much extended reasoning to apply. New effort controls give developers explicit control over the intelligence-speed-cost tradeoff. If the model appears to be overthinking on simpler tasks, users can dial effort down from the default high setting to medium using the /effort parameter.

Office integration received significant upgrades as well. Claude in Excel has been substantially improved, and Claude in PowerPoint is now available in research preview. These enhancements position Anthropic to compete directly with Microsoft’s Copilot and Google’s Workspace AI offerings.

Safety at the Frontier

Anthropic emphasized safety alongside capability improvements. According to the company’s published system card, Opus 4.6 shows an overall safety profile as good as or better than any other frontier model, with low rates of misaligned behavior across safety evaluations.

The model’s pricing remains unchanged at $5 per million input tokens and $25 per million output tokens—positioning it as a premium offering compared to competitors. It’s available immediately on claude.ai, the Anthropic API, and all major cloud platforms.

For developers, the model identifier is claude-opus-4-6. The release maintains Anthropic’s roughly four-month cadence for major model updates, signaling the company has no intention of slowing its development pace despite intensifying competition from OpenAI, Google, and emerging challengers.

“Claude Opus 4.6 represents a meaningful leap in long-context performance. In our testing, we saw it handle much larger bodies of information with a level of consistency that strengthens how we design and deploy complex research workflows.” — Research Partner

The Road Ahead

The release tightens an already competitive race at the frontier of AI capabilities. OpenAI’s GPT-5.2, released earlier this year, had established itself as the benchmark for complex reasoning tasks. Opus 4.6 challenges that position directly, particularly in coding and professional knowledge work.

Industry observers are watching to see how the market responds. Enterprise customers—who represent the most lucrative segment of the AI market—are increasingly sophisticated in their evaluations. They care less about benchmark scores and more about reliable performance on specific workflows. Early partner feedback suggests Opus 4.6 delivers on that front.

The question now is whether Anthropic can convert technical superiority into market share. The company has historically prioritized safety and capability over aggressive commercialization. But with each release, the gap between research achievement and business impact narrows.

This article was reported by the ArtificialDaily editorial team. For more information, visit Anthropic.