Claude Opus 4.6 Announcement

Overview

Anthropic has unveiled Claude Opus 4.6, an enhanced version of its most capable model. The update brings significant improvements in coding abilities, agentic task execution, and context window capacity.

Key Features

Enhanced Capabilities:

• Improved planning and execution for agentic tasks
• Better code review and debugging functionality
• 1M token context window in beta (first for Opus-class models)
• Up to 128k output token support

Performance Highlights:

• Highest score on Terminal-Bench 2.0 for agentic coding
• Leads on Humanity's Last Exam for multidisciplinary reasoning
• Outperforms GPT-5.2 by approximately 144 Elo points on GDPval-AA evaluation
• Best performance on BrowseComp for locating hard-to-find information

Product Updates

Claude Code:

• New agent teams feature for parallel task coordination
• Context compaction for longer-running operations

Office Integration:

• Substantial Excel improvements with better planning and multi-step handling
• Claude in PowerPoint available in research preview

API Improvements:

• Adaptive thinking for contextual extended reasoning
• Effort controls (low, medium, high, max) for intelligence/speed/cost optimization
• US-only inference option available

Safety Profile

Opus 4.6 maintains strong safety standards with "low rates of misaligned behavior" and comprehensive evaluations including new cybersecurity probes. The model shows alignment comparable to or better than other frontier models.

Availability & Pricing

Available on claude.ai, the Claude API, and major cloud platforms. Pricing remains $5/$25 per million tokens; premium pricing applies for prompts exceeding 200k tokens on the Claude Platform.