Anthropic introduced Claude Opus 5 on July 24, 2026, describing it as its strongest Opus model for complex everyday tasks, programming, research and office work. The company says the model approaches the performance of its most powerful system, Claude Fable 5, while completing many tasks at roughly half the cost.
According to Anthropic, the distinguishing feature of Opus 5 is not only greater intelligence but a better ability to see tasks through. The company says the model is improved at reviewing its own work, correcting errors when they surface and continuing to work on a hard problem until it reaches a satisfactory result.
Coding and agentic work
Anthropic singled out programming as a particular strength. In some evaluations, the company says, Claude Opus 5 outperformed other leading models, and in one test its results came very close to Claude Fable 5 at about half the cost of completing the task. Anthropic also reports a notable improvement over its predecessor, Claude Opus 4.8.
For office and business tasks, the model has been made more autonomous. Anthropic says it can understand a complex problem and work through it in several stages, change its approach when necessary and check its own mistakes rather than stopping at the first result. In one example provided by the company, the model fixed a difficult programming issue by locating the root cause rather than patching the visible symptom.
Research and safety
Anthropic also claims gains in scientific research, saying Claude Opus 5 beat Opus 4.8 on several evaluations in biology, organic chemistry and protein-related tasks. The company acknowledged that limits remain in highly complex, long-running autonomous scientific research.
On safety, Anthropic describes Opus 5 as among its safer and more instruction-following models to date. In cybersecurity, the model can help find software vulnerabilities, but Anthropic says tighter restrictions apply to capabilities that would allow those vulnerabilities to be exploited in harmful attacks.
Pricing and availability
Claude Opus 5 is priced at $5 per million input tokens and $25 per million output tokens, the same as Claude Opus 4.8. A faster mode offers roughly two and a half times the speed at double the base price.
The model is available now to consumers. It has become the new default model in Claude Max and is offered as the most capable available model in Claude Pro. Developers can access it through the Claude API.
With Opus 5, Anthropic is aiming not simply for the most powerful AI but for a better balance among intelligence, price and everyday usability. If the company’s reported results hold up in real-world use, the model could matter most to users who want high-end AI capability without paying for the most expensive flagship model on every task.
