4 min readNew DelhiUpdated: Feb 6, 2026 08:40 AM IST Even as the world is coming to terms with Anthropic’s recent product enhancements that triggered a global stock selloff across sectors, the AI company has released an update to its smartest model. The newly launched Claude Opus 4.6 is far ahead of its predecessor in coding skills. For the first time, Opus 4.6 also introduces a 1-million-token context window in beta.
For the uninitiated, a one million context window here means that models like Claude Opus 4.6 can remember and consider up to one million tokens of text at once when replying. Tokens are small units of data that include words, parts of words, or characters that let AI process massive documents, full books, or big projects in a single pass instead of breaking them up. In essence, one million context window here is like offering a model with a huge working memory to keep far more information in view when thinking and answering.
The new update brings the ability to plan more carefully, sustained agentic tasks for longer, the capability to operate more reliably in larger codebases, enhanced code review and debugging skills to detect its own mistakes. Claude Opus 4.6 sustains agentic tasks for longer, an ability that all AI models are aiming to achieve.
When it comes to use cases, Anthropic said that Claude Opus 4.6 comes with some improved capabilities that allow it to execute a wide range of everyday tasks such as running financial analyses, doing research, and using and creating documents, spreadsheets, and presentations. Besides, inside Claude Cowork, where the chatbot can work autonomously, Opus 4.6 can put all the skills to work on a user’s behalf.
Performance benchmarks
Claude Opus 4.6 achieved the highest score on the agentic coding evaluation Terminal-Bench 2.0, and it leads in the complex multidisciplinary reasoning test Humanity’s Last Exam. When it comes to finance, legal and other domains, Opus 4.6 has outperformed OpenAI’s GPT-5.2 and its own Opus 4.5. On the test about the ability to find hard-to-locate information online – BrowseComp – Opus 4.6 has outperformed other frontier models.
When it comes to safety and security, Opus 4.6 demonstrates low rates of misaligned behaviour like deception. It reportedly exhibits fewer unnecessary refusals. Besides, the company has also introduced new cybersecurity measures to evaluate both defensive and offensive security potential.
The Claude Opus 4.5 is now available via the Anthropic API on Claude.ai and across major Claude platforms. Key product integrations include the ‘agent teams’ feature that allows for parallel work on large codebase reviews through Claude Code; autonomous multi-step task execution through Cowork Environment; and Office Suite.
Moving towards AGIStory continues below this ad
The latest update from Anthropic intensifies its competition with OpenAI and Google, by showcasing state-of-the-art performance on key benchmarks for economically valuable knowledge work and agentic coding.
Earlier this week, Anthropic shook the tech sector after its recent product updates, including a legal tool that caused major market shifts, with nearly a trillion dollars being wiped out in tech stocks. Anthropic and other AI labs seem to be pursuing ‘more vertical sections of the labour force’.
The Dario Amodei-led company started off by offering some of the most cutting-edge coding models that caught on with software engineers. Now, they are expanding into other sectors like financial services, legal, and healthcare. And the chaos in the last few days is evidence of how reactive the global market is to any technological shift related to these sectors. Most AI companies are releasing advanced AI models successively in line with their ultimate goal of achieving artificial general intelligence, or AGI – a super capable AI that can replace all intellectual work.