Anthropic Launches Claude Opus 4.6 with Enhanced Capabilities and Safety

New Model Overview

Anthropic introduced Claude Opus 4.6, describing it as a direct upgrade from the prior version. The company highlighted the model’s ability to tackle complex, multi‑step tasks and deliver production‑ready quality on the first attempt, reducing the need for iterative revisions in documents, spreadsheets, and presentations.

Enhanced Knowledge‑Work Features

Beyond coding, Opus 4.6 is optimized for a wider set of knowledge‑work activities. Anthropic emphasized improvements in creating PowerPoint presentations, Excel documents, and advanced search functions. The release is tied to the company’s Cowork product, a non‑technical‑worker‑friendly version of Claude Code, intended to broaden use cases in research, marketing, and other non‑technical domains.

Developer‑Focused Agent Teams

The update adds a research‑preview feature called “agent teams,” allowing the model to operate within Claude Code similarly to a real engineering team. This enables splitting a project across multiple agents, each handling a portion of the work and coordinating with one another. The model also offers a one‑million token context window in beta, responding to user feedback that longer context was a key request.

Safety and Security Testing

Anthropic ran its most comprehensive safety suite for Opus 4.6 to date. New evaluations target user well‑being, the model’s ability to refuse potentially dangerous requests, and its capacity to avoid covert harmful actions. Six new cybersecurity probes were added to monitor misuse, reflecting heightened security capabilities.

Pricing and Availability

Claude Opus 4.6 is available immediately with the same pricing as its predecessor. Anthropic positions the model as a step toward broader adoption of AI across varied knowledge‑work tasks, aiming to expand the hype beyond coding and capture a larger market share.