Anthropic’s New AI Model Targets Coding, Enterprise Work
- By John K. Waters
- 02/17/26
Anthropic has actually launched Claude Opus 4.6, presenting a million-token context window and automated agent coordination features as the AI company looks for to broaden beyond software application development into broader business applications.
The San Francisco-based company stated the model enhances efficiency on coding tasks, monetary analysis, and file processing compared to its predecessor. Anthropic positioned the release as enhancing its position in enterprise AI workflows, a progressively congested market where it completes directly with OpenAI and Google.
“We’re focused on constructing the most capable, trustworthy, and safe AI systems,” an Anthropic spokesperson said. “Opus 4.6 is even better at planning, assisting solve the most intricate coding jobs.”
The release comes 3 days after OpenAI released a desktop application for its Codex AI coding system, underscoring the fast rate of competition in AI development tools. Anthropic said in November that Claude Code, its coding product, reached $1 billion in annualized earnings six months after general accessibility.
Extended Context and Representative Coordination
Opus 4.6 supports up to one million tokens of context in beta on Anthropic’s designer platform, a substantial increase from the 200,000-token limitation of earlier Opus variations. The growth enables the design to process larger codebases and longer documents without splitting tasks across multiple requests.
The business also presented representative groups in Claude Code as a research sneak peek, permitting multiple AI representatives to work simultaneously on segmented parts of a job. Scott White, Anthropic’s head of product, compared the feature to coordinating a human team operating in parallel.
Anthropic stated Opus 4.6 addresses context degradation, a typical issue where AI performance decreases as discussions extend. On a retrieval criteria that hides details in big text volumes, Opus 4.6 scored 76% compared to 18.5% for its Sonnet 4.5 model.
The design supports outputs of as much as 128,000 tokens. Anthropic presented adaptive thinking, which allows the model to figure out when to apply deeper thinking, and four effort settings that developers can adjust to stabilize efficiency, speed, and expense.
Criteria Performance
Anthropic reported that Opus 4.6 leads on Terminal-Bench 2.0, an assessment of AI agents completing command-line jobs, with a 65.4% rating under maximum-effort settings. The Terminal-Bench project’s public leaderboard reveals different entries for Opus 4.6, with a score of 62.9% under one setup.
On GDPval-AA, a benchmark determining performance on expert tasks across finance, legal, and other domains, Anthropic stated Opus 4.6 outperforms OpenAI’s GPT-5.2 by around 144 Elo points, a gap that corresponds to an approximately 70% win rate in direct contrasts. Artificial Analysis, which preserves the GDPval-AA leaderboard, describes the evaluation framework in its method documentation.
Anthropic also pointed out arise from BrowseComp, an OpenAI standard for searching representatives that determines the ability to locate hard-to-find details throughout 1,266 questions that require persistent web navigation.
Security Screening and Cybersecurity Measures
Anthropic said Opus 4.6 went through comprehensive safety assessments, including tests for deception, sycophancy, and cooperation with potential abuse. The company’s system card reports the model revealed low rates of problematic habits while attaining the most affordable rate of over-refusals amongst recent Claude models.
The business developed 6 cybersecurity probes to identify harmful uses of the model’s boosted abilities. Anthropic said it is utilizing Opus 4.6 to determine and patch vulnerabilities in open-source software as part of protective cybersecurity efforts.
“Agents have tremendous capacity for positive impacts in work, but it’s important that representatives continue to be safe, dependable, and trustworthy,” the spokesperson stated, referring to a framework Anthropic released detailing core concepts for agent advancement.
Item Combinations and Prices
Anthropic released Claude in PowerPoint as a research study preview for paid subscribers, constructing on existing integrations with Excel. The PowerPoint tool reads layouts, fonts, and slide design templates to produce discussions, the company stated.
White stated Anthropic has actually observed using Claude Code expanding beyond software engineers to item managers, monetary experts, and employees in other fields. The business pointed out releases at Uber, Salesforce, Accenture, Spotify, and other enterprises.
Opus 4.6 is readily available on claude.ai and through the Claude API under the identifier claude-opus-4-6. Rates stays $5 per million input tokens and $25 per million output tokens. Premium rates of $10 per million input tokens and $37.50 per million output tokens applies when prompts exceed 200,000 tokens utilizing the million-token context window. The design is also available through Amazon Bedrock and Google Cloud Vertex AI.
The release arrives as OpenAI’s GPT-5.3-Codex started rolling out through GitHub Copilot, according to GitHub’s changelog. GitHub described GPT-5.3-Codex as OpenAI’s most current agentic coding model and outlined availability for Copilot Pro, Company, and Enterprise users.
For more information, visit the Anthropic website.
About the Author John K. Waters is the editorial director of a variety of Converge360.com websites, with a concentrate on high-end advancement, AI and future tech. He’s been discussing advanced technologies and culture of Silicon Valley for more than 20 years, and he’s written 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 safeguarded]