Anthropic Announces Claude Opus 4.5: The First AI Model to Break the 80% Programming Benchmark, Outperforming Humans and Reducing Price by 67%
News Summary
Anthropic officially released its latest flagship AI model, Claude Opus 4.5, on November 24, 2025. This model achieves breakthrough advancements in software development, reasoning capabilities, and complex multi-step task processing. It is now available to users via the Claude application, API, and three major cloud platforms. This marks Anthropic's third major model release within two months, following Sonnet 4.5 in September and Haiku 4.5 in October, signaling that AI industry competition has entered a white-hot stage.
The most significant change with Claude Opus 4.5 is a substantial price reduction. It is priced at $5 per million input tokens and $25 per million output tokens, representing a 67% decrease compared to its predecessor, Opus 4.1's $15/$75. This makes top-tier AI capabilities more accessible. This pricing strategy makes it more attractive in competition with OpenAI's GPT-5.1 ($1.25/$10) and Google's Gemini 3 Pro ($2/$12).
In terms of performance, Claude Opus 4.5 achieved an accuracy rate of 80.9% on the SWE-bench Verified benchmark, becoming the first model to break the 80% barrier. It surpassed OpenAI's GPT-5.1-Codex-Max (77.9%) and Google's Gemini 3 Pro (76.2%). This benchmark specifically tests AI systems' performance in real-world software engineering tasks, and Claude Opus 4.5's score represents a new industry benchmark.
Even more astonishingly, Anthropic tested Opus 4.5 using the actual technical exam given to candidates for performance engineer roles at the company. The model's score exceeded the highest historical scores of all human applicants. This result has sparked in-depth discussions within the industry about how AI technology will reshape white-collar professions.
Technically, Claude Opus 4.5 features a 200,000-token context window and a 64,000-token output limit, with its knowledge cutoff updated to March 2025. The model has undergone significant improvements in memory management, specifically optimized for long-context operations, enabling it to intelligently recall key details. These enhancements make it particularly suitable as a primary agent to orchestrate collaborative work scenarios involving multiple Haiku sub-agents.
In practical applications, early testers reported that Opus 4.5 could handle tasks that Sonnet 4.5 found almost impossible, finding solutions to complex multi-system problems without hands-on guidance. Renowned developer Simon Willison used Claude Code over the weekend to perform a large-scale refactoring of sqlite-utils, completing 20 commits involving 39 files, 2022 lines of new code, and 1173 lines deleted within two days.
Regarding safety, Anthropic states that Opus 4.5 is its most robustly aligned model to date, achieving significant progress in resisting prompt injection attacks, making it harder to trick than any other cutting-edge model in the industry. This is particularly crucial for enterprise customers using Claude for critical tasks.
Synchronized with the model release, Anthropic introduced a series of product updates: the Claude for Chrome extension is now available to all Max users, Claude for Excel officially launched for Max, Team, and Enterprise users, supporting pivot tables, charts, and file uploads. The desktop version of the Claude Code application has also been officially released, supporting Windows, macOS, and Windows (Arm 64) platforms, allowing developers to run multiple coding or research sessions in parallel.
Notably, Anthropic has adjusted usage limits, allowing users with access to Opus 4.5 to use the model at levels similar to previous Sonnet tiers. This means users don't need to worry about excessive restrictions in their daily work.
In terms of market competition, Microsoft and Nvidia announced multi-billion dollar investments in Anthropic last week, boosting the AI lab's valuation to approximately $350 billion. Anthropic achieved an annualized revenue of $2 billion in Q1 2025, doubling from $1 billion in the previous quarter, with an 8x year-over-year increase in customers spending over $100,000 annually.
The release of Claude Opus 4.5 comes amidst intense competition in the AI industry. OpenAI released GPT-5.1 on November 12, Google launched Gemini 3 on November 18, and now Anthropic responds to the market with Opus 4.5. Anthropic Product Lead Scott White stated: "I'm incredibly excited about the volume of products we're releasing to market and the feedback loop that's generating."
Regarding target user groups, White noted that Opus 4.5's ideal users are professional software developers and knowledge workers, such as financial analysts, consultants, and accountants, as well as those eager to drive creativity and build new things.
Developers can call Claude Opus 4.5 via API using the model string "claude-opus-4-5-20251101." They can also enjoy 90% cost savings from prompt caching and 50% cost savings from batch processing. The model is already available on platforms such as Amazon Bedrock, Google Cloud's Vertex AI, and Microsoft Foundry.
Anthropic emphasizes that Opus 4.5 is an advanced model designed for "unprecedented use cases," particularly suitable for professional software engineering, complex agent workflows, and high-stakes enterprise tasks. Its hybrid reasoning capabilities allow flexible switching between immediate responses and extended deliberation. API users can finely tune the overall effort the model invests in its responses, balancing performance, latency, and cost.
Feedback from industry partners also confirms Opus 4.5's powerful capabilities. Lovable stated that the model provides cutting-edge reasoning capabilities in its chat mode, with deep reasoning transforming planning, and excellent planning leading to better code generation. Warp reported a 15% improvement with Opus 4.5 over Sonnet 4.5 in Terminal Bench tests, particularly noticeable when using Planning Mode. Nico Christie, co-founder of financial modeling firm Fundamental Research Labs, said that internal evaluations showed a 20% increase in accuracy and a 15% improvement in efficiency, making previously seemingly unattainable complex tasks now achievable.
The release of Claude Opus 4.5 not only represents a new breakthrough in Anthropic's technological prowess but also signals that AI assistants are evolving from simple Q&A tools into intelligent systems capable of independently completing complex professional tasks. With significant price reductions and substantial capability enhancements, the commercial application of AI technology is expected to accelerate further, bringing profound changes to various industries.