Artificial Analysis has released version 2.1 of its Coding Agent Index, in which Claude Fable 5.2 and Claude Opus 5.5 share first place with a composite score of 58. The update replaces FrontierCode v1.1 and Terminal-Bench 4.0 and covers approximately 290 tasks weighted equally across three benchmarks: FrontierCode, Terminal-Bench and SWE-Atlas-QnA-AA. The new version evaluates each model together with its agent harness, while Claude Opus 5.5’s score fell from 61.7 in version 2.0 to 58 because of the test changes. The decline therefore does not necessarily indicate worse performance, but rather a different measurement framework. These results primarily concern the evaluation of software development capabilities and do not provide a direct read-through to financial markets.
Source: Read the original article
