Anthropic introduces Claude Opus 5: a new benchmark in coding and intellectual work at half the price

Anthropic has officially announced the release of its new language model, Claude Opus 5. According to the developers, the model demonstrates a superior level of performance, outperforming existing solutions in programming tasks and complex intellectual work.

While matching the efficiency of the Fable 5 model, the new offering from Anthropic costs half as much, making it one of the most cost-effective solutions on the market.

A new level of autonomy and code handling

In programming benchmarks, Claude Opus 5 showed results unattainable by other models with similar configurations. Anthropic representatives shared real-world use cases for the AI:

1. Creating 3D models from pixels (computer vision)

As part of an experiment, the model was given a blueprint of a machine part with the task of writing code to reconstruct it as a 3D model in FreeCAD. Opus 5 did not have direct access to view the blueprint. To solve the task, the model independently developed a computer vision pipeline, extracted geometry directly from the raw pixels, and fully reconstructed the part. Opus 5 successfully repeated this process several times, while none of the competing models managed to complete the task even after five attempts.

2. Finding and fixing deep-seated bugs

Opus 5 was tasked with analyzing the code of a popular messenger app. The model not only discovered the bug and identified its root cause but also fixed a specific edge case that community developers had missed when creating the official patch. A competing model in the same test only addressed the superficial symptoms, erroneously reporting a complete fix.

3. Financial infrastructure development and auto-testing

A trading company engineer used Opus 5 to write a market data feed for a new exchange. The task was completed in a single session. Previously, no model could solve this problem even with detailed instructions and plans provided by the engineer. Without access to a real data feed to verify the code, Opus 5 independently created a test environment and verified the correctness of the data processing.

Conclusion

Claude Opus 5 sets a new bar for AI assistants in the development field. The model demonstrates the ability not just to write code based on instructions, but to create auxiliary tools (vision pipelines, test environments) to verify its own solutions.