Anthropic Introduces Claude Opus 5, Its New Flagship AI Model

Anthropic Introduces Claude Opus 5, Its New Flagship AI Model

Quick Reads

• Anthropic says Claude Opus 5 delivers near-frontier intelligence at half the cost of Claude Fable 5.
• The model outperformed other systems on coding, knowledge work, automation, and computer-use benchmarks.
• Opus 5 achieved three times the score of the next-best model on the ARC-AGI 3 evaluation.
• Anthropic has priced Opus 5 at $5 per million input tokens and $25 per million output tokens.

Anthropic has launched Claude Opus 5, a new flagship AI model that aims to deliver near-frontier intelligence at a lower cost. The company says Opus 5 comes close to the performance of Claude Fable 5 while costing half as much. Anthropic says the model has achieved state-of-the-art results on coding and knowledge work evaluations. The launch adds fresh competition to an AI market where companies are racing to deliver more capable models without sharply increasing costs for developers and businesses.

Stronger Performance Across Coding and Automation

The model’s strongest gains appear in software engineering, automation, and complex problem-solving. On Frontier-Bench v0.1, Anthropic says Opus 5 outperformed every other model and more than doubled the performance of its predecessor, Opus 4.8. Meanwhile, the model came within 0.5% of Fable 5’s peak score on CursorBench 3.2 at maximum effort, while costing half as much per task. On ARC-AGI 3, which tests how well AI systems solve novel problems, Opus 5 achieved three times the score of the next-best model. In addition, it recorded strong results on business automation and computer-use evaluations, potentially making it useful for developers and companies handling complex digital workflows.

Anthropic also highlighted Claude Opus 5’s performance in scientific research. The model showed strong results in life sciences. These include structural biology, organic chemistry, and bioinformatics. For example, Opus 5 scored 10.2 percentage points higher on an internal organic chemistry benchmark. The test involved identifying molecular structures from spectroscopy data. It also improved by 7.7 percentage points on tasks involving protein sequences and their functions. Beyond research, Anthropic says Opus 5 can check its work and make careful improvements. In early tests, the model reconstructed a 3D machine part from raw image pixels. It also fixed a difficult software bug that another model failed to resolve.

The launch also comes with a focus on safety and cost efficiency. Anthropic says Opus 5 is its most aligned model to date and remains behind its Mythos 5 model on biology and offensive cybersecurity tasks. However, the company said the model can identify software vulnerabilities while remaining less capable of developing exploits. Developers can access Opus 5 through the Claude API at $5 per million input tokens and $25 per million output tokens, the same pricing as Opus 4.8. As a result, the model could make advanced AI more accessible for coding, research, automation, and enterprise work. Looking ahead, Claude Opus 5’s combination of performance, lower cost, and broader capabilities could intensify competition among AI companies seeking to make powerful models practical for everyday use.