Anthropic has introduced Claude Opus 4.5, its latest and most powerful artificial intelligence model, marking a significant milestone in AI software engineering capabilities. Unveiled on Monday, this new model not only achieves superior performance on complex coding tasks but also dramatically lowers costs by approximately two-thirds compared to its predecessor, Claude Opus 4.1.
Backed by Amazon, Anthropic is offering Claude Opus 4.5 at $5 per million input tokens and $25 per million output tokens, a steep price reduction from the earlier $15 and $75 rates. This pricing strategy aims to broaden access to cutting-edge AI technology for developers and enterprises, while increasing competitive pressure on industry leaders like OpenAI and Google.
Superior Performance in Engineering Assessments
Claude Opus 4.5 achieved an 80.9% accuracy rate on SWE-bench Verified, a benchmark assessing practical software engineering tasks. This score outperforms OpenAI’s GPT-5.1 Codex Max at 77.9%, Anthropic’s own Sonnet 4.5 at 77.2%, and Google’s Gemini 3 Pro at 76.2%. Notably, the model scored higher than any human candidate in Anthropic’s most rigorous internal engineering exam, which tests both technical skills and judgment under time constraints.
Alex Albert, Head of Developer Relations at Anthropic, emphasized the model’s enhanced judgment and intuition, describing it as a qualitative leap in understanding real-world contexts. He noted that Claude Opus 4.5 can autonomously synthesize and prioritize information more effectively than previous models, significantly improving workflow integration.
Efficiency Gains and Customizable Effort Parameters
Beyond raw performance, Claude Opus 4.5 features substantial efficiency improvements, reducing token consumption by up to 76% on key benchmarks without sacrificing quality. At medium effort settings, it matches Sonnet 4.5’s best scores while using 76% fewer output tokens, and at the highest effort level, it surpasses Sonnet 4.5’s performance by 4.3 percentage points with 48% fewer tokens.
Anthropic introduced an “effort parameter” allowing users to balance computational intensity against latency and cost, giving developers greater control over AI resource usage. Early adopters such as Replit and GitHub have reported significant efficiency and performance benefits, especially in tasks like code migration and refactoring.
Self-Improving AI Agents and Expanded Capabilities
Early customers have highlighted the model’s ability to act as a “self-improving agent,” iteratively refining its performance through repeated execution without altering its core parameters. Rakuten’s AI team reported peak task performance within four iterations, compared to ten iterations for other models.
These capabilities extend beyond coding to include the generation of professional documents, spreadsheets, and presentations. According to Anthropic, this generational leap from Sonnet 4.5 to Opus 4.5 represents the biggest improvement observed between successive models to date.
New Features Enhance User Experience and Integration
Complementing the model release, Anthropic launched several product enhancements targeting enterprise users. Claude for Excel now supports pivot tables, charts, and file uploads and is available to Max, Team, and Enterprise customers. A Chrome browser extension has been made accessible to Max users.
Significantly, Anthropic introduced “infinite chats,” a feature that overcomes traditional context window limitations by summarizing earlier conversation content dynamically, effectively providing an unlimited context window. Developers also gained access to “programmatic tool calling,” enabling Claude to write and execute code that directly invokes functions. Additionally, Claude Code now includes an updated “Plan Mode” and supports parallel AI agent sessions on desktop in research preview.
Market Dynamics and Industry Competition
Anthropic’s release of Opus 4.5 comes amid fierce competition with OpenAI and Google, both of which have recently launched new AI models and variants. OpenAI’s GPT-5.1 and Codex Max, which can operate autonomously for extended periods, and Google’s Gemini 3 released in November, underscore the escalating innovation race.
Anthropic reported $2 billion in annualized revenue for Q1 2025, doubling from the previous quarter, with a significant increase in customers spending over $100,000 annually. The company credits its accelerated development pace partly to leveraging Claude AI internally for product and research tasks.
Despite reduced pricing potentially impacting margins, Anthropic anticipates wider adoption among startups integrating advanced AI features into their products. Industry experts recognize Opus 4.5 as a substantial improvement in coding intelligence and cost-efficiency, with implications for the future of professional work as AI performance approaches or exceeds human expert levels.
Reflecting on the model’s achievement in surpassing human candidates on the engineering exam, Albert remarked, “I think it’s a really important signal to pay attention to,” highlighting the transformative potential of AI in workplace productivity and professional roles.
Fonte: ver artigo original
Related Articles
Cloudflare Strengthens AI Capabilities with Acquisition of Replicate and Its Vast Model Library
Former OpenAI Policy Chief Launches Institute for Independent AI Safety Audits
Ocorian Study Reveals Family Offices Increasingly Rely on AI for Financial Data Insights
Google Cloud and DeepMind Deploy Gemini AI to Automate UK Council Planning
