Anthropic’s New AI Design Targets Coding, Business Work
- By John K. Waters
- 02/17/26
Anthropic has launched Claude Opus 4.6, introducing a million-token context window and automated agent coordination includes as the AI company seeks to expand beyond software application advancement into wider enterprise applications.
The San Francisco-based firm stated the model enhances performance on coding jobs, monetary analysis, and document processing compared to its predecessor. Anthropic placed the release as reinforcing its position in enterprise AI workflows, a significantly crowded market where it completes straight with OpenAI and Google.
“We’re focused on building the most capable, reputable, and safe AI systems,” an Anthropic spokesperson said. “Opus 4.6 is even better at planning, assisting fix the most complicated coding jobs.”
The release comes 3 days after OpenAI launched a desktop application for its Codex AI coding system, highlighting the rapid speed of competition in AI advancement tools. Anthropic stated in November that Claude Code, its coding product, reached $1 billion in annualized profits 6 months after general availability.
Extended Context and Representative Coordination
Opus 4.6 supports approximately one million tokens of context in beta on Anthropic’s designer platform, a considerable increase from the 200,000-token limit of earlier Opus versions. The growth allows the design to process larger codebases and longer files without splitting jobs throughout numerous demands.
The business also presented agent teams in Claude Code as a research study sneak peek, allowing several AI agents to work at the same time on segmented portions of a project. Scott White, Anthropic’s head of product, compared the feature to collaborating a human team operating in parallel.
Anthropic said Opus 4.6 addresses context destruction, a typical problem where AI efficiency declines as discussions extend. On a retrieval standard that conceals details in large text volumes, Opus 4.6 scored 76% compared to 18.5% for its Sonnet 4.5 design.
The design supports outputs of as much as 128,000 tokens. Anthropic introduced adaptive thinking, which allows the model to identify when to use much deeper reasoning, and four effort settings that developers can adapt to stabilize efficiency, speed, and cost.
Criteria Efficiency
Anthropic reported that Opus 4.6 leads on Terminal-Bench 2.0, an assessment of AI agents finishing command-line tasks, with a 65.4% score under maximum-effort settings. The Terminal-Bench task’s public leaderboard reveals separate entries for Opus 4.6, with a score of 62.9% under one setup.
On GDPval-AA, a benchmark measuring performance on expert tasks across financing, legal, and other domains, Anthropic said Opus 4.6 surpasses OpenAI’s GPT-5.2 by approximately 144 Elo points, a gap that corresponds to an approximately 70% win rate in direct contrasts. Synthetic Analysis, which preserves the GDPval-AA leaderboard, explains the evaluation structure in its method documentation.
Anthropic likewise pointed out arise from BrowseComp, an OpenAI standard for browsing representatives that measures the ability to find hard-to-find details throughout 1,266 questions that require relentless web navigation.
Safety Screening and Cybersecurity Measures
Anthropic said Opus 4.6 went through substantial safety assessments, consisting of tests for deceptiveness, sycophancy, and cooperation with potential misuse. The business’s system card reports the design showed low rates of troublesome habits while accomplishing the most affordable rate of over-refusals amongst current Claude models.
The company developed 6 cybersecurity probes to identify harmful uses of the design’s enhanced capabilities. Anthropic stated it is utilizing Opus 4.6 to recognize and spot vulnerabilities in open-source software application as part of defensive cybersecurity efforts.
“Agents have incredible potential for positive effects in work, however it is necessary that agents continue to be safe, trusted, and trustworthy,” the representative stated, describing a framework Anthropic released detailing core concepts for representative advancement.
Product Combinations and Prices
Anthropic released Claude in PowerPoint as a research preview for paid subscribers, constructing on existing integrations with Excel. The PowerPoint tool checks out designs, fonts, and slide templates to generate discussions, the company said.
White stated Anthropic has actually observed the use of Claude Code expanding beyond software engineers to item managers, financial analysts, and workers in other fields. The company cited releases at Uber, Salesforce, Accenture, Spotify, and other enterprises.
Opus 4.6 is offered on claude.ai and through the Claude API under the identifier claude-opus-4-6. Pricing stays $5 per million input tokens and $25 per million output tokens. Premium pricing of $10 per million input tokens and $37.50 per million output tokens uses when triggers exceed 200,000 tokens using the million-token context window. The model is likewise offered through Amazon Bedrock and Google Cloud Vertex AI.
The release shows up as OpenAI’s GPT-5.3-Codex started presenting through GitHub Copilot, according to GitHub’s changelog. GitHub explained GPT-5.3-Codex as OpenAI’s most current agentic coding design and detailed schedule for Copilot Pro, Business, and Business users.
For more details, go to the Anthropic website.
About the Author John K. Waters is the editorial director of a number of Converge360.com websites, with a focus on high-end advancement, AI and future tech. He’s been writing about innovative innovations and culture of Silicon Valley for more than 20 years, and he’s composed more than a dozen books. He also co-scripted the documentary Silicon Valley: A 100 Year Renaissance, which aired on PBS. He can be reached at [e-mail secured]