Upgrading the model ahead of GPT-5, Anthropic releases Opus 4.1, with enhanced capabilities in programming, research, and data analysis

The competition among artificial intelligence (AI) models is heating up again. As OpenAI is about to release the highly anticipated GPT-5, Anthropic has taken the lead by upgrading its own model, launching Claude Opus 4.1, claiming significant improvements in programming, research, and data analysis capabilities.

On August 5th, Tuesday, Eastern Time, Anthropic, founded by former OpenAI employees, announced that the new model Opus 4.1 scored 74.5% on the programming assessment benchmark SWE-Bench Verified, an increase of two percentage points from the previous generation Opus 4's 72.5%.

The new model excels particularly in navigating large codebases and multi-file code refactoring. Feedback from clients such as GitHub and Rakuten Group indicates that Opus 4.1 has shown significant improvements in code modification accuracy and debugging efficiency, being able to accurately locate code that needs fixing without introducing vulnerabilities.

In the face of competitive pressure from OpenAI, which may release GPT-5 this month, Anthropic has chosen to focus on optimizing existing products.

Mike Krieger, Chief Product Officer of Anthropic, stated that this upgrade of the Opus model marks a strategic shift for the company towards more frequent incremental improvements rather than solely focusing on major version updates. He said:

“In the past, we were too focused on only providing significant upgrades. (The model) is now superior in coding, reasoning, and agent tasks. We just want it to serve humanity better.”

Performance Improvements Focused on Programming

Data released by Anthropic shows that Opus 4.1 has achieved substantial breakthroughs in programming capabilities.

Anthropic announced that in the SWE-Bench Verified benchmark test assessing large language models (LLM) for real-world software engineering capabilities, Opus 4.1 achieved an accuracy rate of 74.5%. This result shows significant improvement compared to Claude Sonnet 3.7's 62.3% and Opus 4's 72.5%.

Anthropic emphasized that the upgraded Opus model is more efficient in handling complex multi-step problems, positioning it as a more effective AI agent. The new model is better at navigating large codebases and is more precise in code modifications.

Opus 4.1 also “enhances Claude's in-depth research and data analysis capabilities, especially in detail tracking and agent search.”

On Tuesday, Anthropic stated that Windsurf, an AI programming assistant acquired by Cognition, reported that Opus 4.1 showed a standard deviation improvement in its junior developer benchmark test compared to Opus 4, with performance improvement comparable to the leap from Sonnet 3.7 to Sonnet 4

Customer Feedback Validates Practical Value

Anthropic mentioned on Tuesday that some enterprise customers' usage feedback confirmed the practical improvements of the new model.

For example, Japanese e-commerce giant Rakuten Group found that Opus 4.1 excels at precisely locating the areas that need correction within large codebases, without unnecessary adjustments or introducing vulnerabilities. Rakuten's team tends to leverage this precision of the model for daily debugging tasks.

Windsurf stated that programming tasks were completed faster and with improved quality after using Opus 4.1. GitHub pointed out that Opus 4.1 has improvements over Opus 4 in most functionalities, with particularly significant performance enhancements in multi-file code refactoring.

Strategic Adjustments Amid Intensifying Market Competition

Anthropic's release comes at a time when competition in the AI industry is heating up. Both Google and OpenAI have launched features to help programmers simplify code writing and debugging processes, while OpenAI executives have been promoting the upcoming GPT-5 in public, with reports suggesting that the product may launch this month.

When asked about OpenAI's upcoming products, Mike Krieger stated, "One thing I've learned is that we need to focus on what we have, especially in the rapidly evolving AI field; what others do ultimately depends on them."

Anthropic announced on Tuesday that Opus 4.1 is now available to paid Claude users and can be accessed through Anthropic's API, Amazon Bedrock, and Google Cloud's Vertex AI, with pricing consistent with Opus 4. Anthropic also plans to release more significant model updates in the coming weeks.

Anthropic Reportedly Seeking New Funding That Could Boost Valuation to $170 Billion

Nearly two weeks ago, in mid-July, media reported that Anthropic claimed its annualized revenue quadrupled in the first half of this year, exceeding $4 billion. Its explosive revenue growth has attracted significant interest from some investors, considering a new round of investment at a valuation exceeding $100 billion, nearly doubling from the $58 billion valuation announced four months ago during its last funding round.

Subsequently, media reported after discussions with several Middle Eastern investors that Anthropic's upcoming valuation is closer to $150 billion.

Last week, media reported that Anthropic is in discussions for a new funding round led by Iconiq Capital, aiming to raise $3 billion to $5 billion, which would bring the company's valuation to $170 billion. Other media sources indicated that by the end of July, Anthropic's annualized revenue had increased to approximately $5 billion. The company expects its recurring revenue to potentially reach $9 billion by the end of this year The new financing news from Anthropic highlights the market's extremely high expectations for the future growth of leading AI companies, especially regarding Anthropic's strong monetization capabilities in the field of AI coding