Anthropic unveiled Claude Opus 4.5 on Monday, marking the company’s third major AI model release within a two-month timeframe and escalating competition in the artificial intelligence sector. The flagship model delivers what the company calls “best-in-world” performance for coding, agents, and enterprise workflows, directly challenging recent releases from OpenAI and Google.
Industry-Leading Coding Performance
Claude Opus 4.5 achieved 80.9% accuracy on SWE-bench Verified, a benchmark testing real-world software engineering tasks. This performance surpasses OpenAI’s GPT-5.1-Codex-Max (77.6%), Anthropic’s own Sonnet 4.5 (72%), and Google’s Gemini 3 (76.2%), according to company statistics. GitHub’s integration reveals the model’s practical advantages—in early testing, Opus 4.5 exceeded internal coding benchmarks while cutting token usage in half, making it especially suited for tasks like code migration and refactoring.
Breakthrough in AI Agents and Autonomous Systems
The model represents a significant advancement in self-improving AI agents capable of refining their own capabilities through iterative learning. For office automation tasks, Anthropic’s agents autonomously achieved peak performance in just 4 iterations, while competing models couldn’t match that quality after 10 attempts. Opus 4.5 handles ambiguity and reasons about tradeoffs without hand-holding, with testers reporting that tasks near-impossible for Sonnet 4.5 weeks ago are now within reach.
Enhanced Efficiency and Pricing
Anthropic made Opus-level capabilities more accessible by pricing the model at $5 per million input tokens and $25 per million output tokens. The model demonstrates remarkable efficiency improvements—at medium effort levels, Opus 4.5 matches Sonnet 4.5’s best SWE-bench Verified score while using 76% fewer output tokens. At the highest effort setting, it surpasses Sonnet 4.5’s performance by 8.3 percentage points while consuming 48% fewer tokens.
Advanced Features and Enterprise Integration
Claude Opus 4.5 introduces an “effort parameter” that defaults to high but can be adjusted to medium or low for faster responses, giving developers granular control over performance versus speed tradeoffs. The model excels at managing teams of subagents for complex multi-agent systems, with advanced context management and memory capabilities boosting performance on agentic tasks by almost 15 percentage points. The model is now available across Anthropic’s apps, API, and all three major cloud platforms, with integrations already rolled out to GitHub Copilot, Chrome, and Excel. Industry partners including Cursor’s CEO Michael Truell praised Opus 4.5 as “a notable enhancement” with improved pricing and intelligence on challenging coding tasks.
