{"id":94013,"date":"2026-02-06T14:49:44","date_gmt":"2026-02-06T11:49:44","guid":{"rendered":"https:\/\/forklog.com\/en\/?p=94013"},"modified":"2026-02-06T14:50:21","modified_gmt":"2026-02-06T11:50:21","slug":"claude-opus-4-6-surpasses-gpt-5-2-in-logic-tests-and-introduces-agent-teams","status":"publish","type":"post","link":"https:\/\/forklog.com\/en\/claude-opus-4-6-surpasses-gpt-5-2-in-logic-tests-and-introduces-agent-teams\/","title":{"rendered":"Claude Opus 4.6 Surpasses GPT-5.2 in Logic Tests and Introduces ‘Agent Teams’"},"content":{"rendered":"

AI startup Anthropic has upgraded its flagship model, Claude Opus, to version 4.6. The neural network has improved its ability to plan actions, handle long-term tasks, and work more efficiently with large codebases.<\/p>\n

The context window has been expanded to 1 million tokens, allowing for the analysis of massive documents and extended dialogues without losing the logical thread.<\/p>\n

The updated algorithms are tailored for professional tasks: conducting financial analysis, research, and the use and creation of documents, spreadsheets, and presentations.<\/p>\n

Opus 4.6 received the highest score in the Terminal-Bench 2.0 programming test and outperformed competitors in the complex interdisciplinary benchmark for logical reasoning, Humanity\u2019s Last Exam.<\/p>\n

\"image\"
Comparison of Opus 4.6 with competitors in various tests. Source: Anthropic.<\/figcaption><\/figure>\n

In GDPval-AA, which assesses reasoning and decision-making quality, the model surpassed OpenAI’s GPT-5.2. LLM<\/span> also achieved better results in BrowseComp, which measures the ability to find hard-to-access information online.<\/p>\n

\"image\"
Source: Anthropic.<\/figcaption><\/figure>\n

Opus 4.6 efficiently extracts data from large documents. Thanks to the expanded context window, the model tracks and captures subtle hidden details.<\/p>\n

Agent Teams<\/h2>\n

A key innovation is the ability to create groups of agents for collaborative work. In this mode, multiple AI assistants work in parallel and coordinate their tasks autonomously.<\/p>\n

This tool is suitable for assignments that are divided into independent parts and require the analysis of large amounts of text.<\/p>\n

Closed Loop<\/h2>\n

Anthropic stated that they are “creating Claude with Claude.” Developers write code using their own AI model, and each new product undergoes testing on internal company tasks before release.<\/p>\n

The team found that Opus 4.6 pays more attention to the most challenging parts of a task without additional instructions, quickly completes simple tasks, handles ambiguous problems better, and maintains efficiency over long distances.<\/p>\n

\n

\u201cOpus 4.6 often thinks more deeply and thoroughly revises its reasoning before making a decision. This yields better results in solving complex cases but may increase costs and expenses in simpler ones,\u201d the company noted.<\/p>\n<\/blockquote>\n

Safety<\/h2>\n

An automated audit revealed that Opus 4.6 has a low propensity for undesirable behavior: deception, flattery, reinforcing user misconceptions, and aiding in improper actions.<\/p>\n

\"image\"
The model demonstrates safety at the level of Opus 4.5. Source: Anthropic.<\/figcaption><\/figure>\n

To evaluate the model, the company conducted the most comprehensive series of assessments, applying new testing methodologies and enhancing existing ones for the first time.<\/p>\n

Availability and New Features<\/h2>\n

Claude Opus 4.6 is now available via web interface, API<\/span>, and on major cloud platforms.<\/p>\n

New features in the developer toolkit include:<\/p>\n