Anthropic launches new Claude model for coding, reasoning tasks

Anthropic has announced the release of Claude Opus 4.6, an upgraded AI model designed for coding, reasoning, and knowledge work.

The new version features a 1 million-token context window in beta, allowing it to handle longer and more complex tasks, including financial analysis and document creation.

It performs well on several industry benchmarks, such as the Terminal-Bench 2.0 coding evaluation and Humanity’s Last Exam reasoning test.

The model outperforms previous versions and competitors like OpenAI’s GPT-5.2 on tasks related to information retrieval and reasoning in long contexts.

Anthropic reports that Opus 4.6 maintains a safety profile comparable to or better than earlier models, with low rates of misaligned behaviors.

The model is available via Anthropic’s API and cloud platforms, with pricing unchanged at US$5 or US$25 per million tokens.

🔗 Source: Anthropic

🧠 Food for thought

Implications, context, and why it matters.

Anthropic’s new AI takes aim at OpenAI for high-value enterprise work

Benchmark results back the performance claims, including stronger recall across long documents.
On a long-context recall test, Opus 4.6 reached 76%, up from 18.5% in the prior model, which makes the 1 million-token context window practical for complex analysis 1.
The release also adds “agent teams,” letting developers split work across multiple AI “agents” that coordinate on complex coding tasks 1.
The launch arrived three days after an OpenAI release. Andreessen Horowitz (a Silicon Valley venture capital firm) survey data cited by VentureBeat puts Anthropic at about 40% of enterprises using it in production by January 2026, up from near-zero in March 2024 2.

AI “agents” are reshaping job roles and rattling software markets

The push toward autonomous work is changing job mix, especially in software development and other professional knowledge work.
Usage clusters in high-skill areas. Computer and mathematical work made up 46% of Anthropic’s first-party API traffic in November 2025 3.
That pattern can drive “deskilling,” where AI handles the hardest parts of a role and people get left with lower-value tasks 3.
Investors have started to price in disruption risk.
Concerns that AI agents could replace established business software fed a $285 billion selloff in software and services stocks. Legal services firms such as Thomson Reuters fell 15.83% in one day 2.