Anthropic Launches Claude 4 Series, Sets New Benchmark with 7-Hour Autonomous Coding

On May 22 (local time), Anthropic, a U.S.-based AI company considered the strongest competitor to OpenAI, officially released its most advanced language model to date—the Claude 4 series.

The flagship model, Claude 4 Opus, has achieved a significant advancement in coding capabilities. According to Rakuten’s test data, programming agents developed with Opus 4 can autonomously and reliably operate continuously for seven hours, surpassing the previous record held by OpenAI.

In addition to excelling in coding tasks, Claude 4 Opus demonstrates improvements in reasoning, planning, and long-term memory. Some experts suggest it may replace certain software engineers as well as junior programmers and code reviewers who focus on simple code writing and maintenance. Nevertheless, Anthropic stresses that its AI models are designed to automate routine work rather than eliminate jobs.

To address varied user needs, Anthropic also introduced Claude Sonnet 4, a more cost-effective variant within the Claude 4 lineup suited for general tasks. Sonnet 4, an iteration of Sonnet 3.7, performs strongly in programming, achieving a 72.7% score on the SWE-bench benchmark, exceeding the performance of OpenAI’s latest models such as Codex - 1 and o3.

Both models are accessible via the Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI.

Regarding safety, the Claude 4 series is classified under ASL-3 security standards. Anthropic has implemented several risk mitigation measures, including a “Thought Summary” feature designed to enhance model interpretability and transparency.

This release represents a key milestone for Anthropic in competing against tech giants such as OpenAI and Google. While OpenAI’s products have wide consumer market reach—ChatGPT attracting a substantial individual user base—Anthropic focuses primarily on the enterprise sector, with annual revenue estimated at $875 million mainly derived from sales of its enterprise product, Claude Enterprise, while continuing to broaden its market presence.

Founded in 2021 by former OpenAI research vice president Dario Amodei and his sister Daniela Amodei, who contributed to the development of OpenAI’s GPT-2 and GPT-3 models, Anthropic was established due to differences regarding AI safety and development priorities. Their goal is to develop AI systems that are safe, reliable, and aligned with human values.

In recent years, Anthropic has secured backing from major tech companies including Amazon and Google. In March 2025, the company completed a $3.5 billion Series E funding round, raising its post-investment valuation to $61.5 billion. Meanwhile, in the same month, OpenAI raised $40 billion in a new funding round led by SoftBank Group, reaching a post-investment valuation of $300 billion, significantly higher than Anthropic’s.

With the introduction of the Claude 4 series, the competitive gap between Anthropic and OpenAI is expected to narrow further.