Anthropic’s Claude Opus 5 delivers a major leap in coding and computer tasks
Anthropic has launched Claude Opus 5, its newest advanced AI model designed for difficult tasks such as coding, research, business operations and computer use.
The company says Opus 5 brings a major improvement over Claude Opus 4.8 while keeping the same API pricing.
Anthropic also claims that Opus 5 can come close to the performance of Claude Fable 5 in some coding and computer-use tests. The key difference is cost, with Opus 5 reportedly completing several tasks much more cheaply.
Claude Opus 5 is now available across Claude applications and through its API. It has also become the default model for Claude Max subscribers and is described as the strongest option available to Claude Pro users.
One of the most notable results came from CursorBench 3.2, a test that measures how well AI coding agents handle real software development work.
At maximum effort, Opus 5 came within 0.5% of Fable 5’s top score. Anthropic says it achieved this result at about half the cost per task.
The model also performed better than Fable 5’s best result on OSWorld 2.0. This test measures how well AI systems can use computers, work with files, and complete tasks across different applications.
Anthropic says Opus 5 achieved the result while costing only a little more than one-third of Fable 5’s price.
This could make Opus 5 especially important for Pro and standard Team subscribers. Fable 5 moved to a pay-as-you-go credit system on July 20, creating a possible gap for users who need a powerful model without higher task costs.
Anthropic also reported a major improvement on Frontier-Bench, which tests AI agents on difficult software engineering problems.
Also Read: PM Shehbaz Sharif launches major robotics initiative with $20 million venture fund
According to the company, Opus 5 more than doubled the performance of Opus 4.8 on this benchmark. It also completed the tasks at a lower average cost.
Another claimed improvement is the model’s ability to check its own work. Anthropic says Opus 5 is better at finding bugs, reviewing its answers and continuing to work through difficult problems instead of quickly giving up.
In one example, Opus 5 reportedly discovered an edge case that had been missed by an earlier community patch.
These results, however, come from Anthropic’s own testing. Actual performance can be different depending on the type of task, software environment and workload.
The pricing has not changed from Claude Opus 4.8. Anthropic lists the API price at $5 per million input tokens and $25 per million output tokens.
