Anthropic’s latest language model, Claude Opus 5.5, is showing a markedly different writing style compared with its predecessor, Opus 5, according to recent analysis. The new version reportedly uses 95 % fewer em dashes, and its responses are longer, yet the sentences are shorter and the vocabulary simpler.
In a study of the two models’ output, researchers noted that Opus 5.5 produces fewer obvious AI‑writing patterns. The reduction in em dashes—a common stylistic marker in machine‑generated text—suggests the model is adopting a more natural, human‑like rhythm in its prose.
Despite the increased length of its answers, Opus 5.5’s sentences are noticeably more concise. The model’s simpler wording may help users read and understand responses more easily, a change that could improve its usability in customer‑facing applications.
Anthropic has not released a formal statement on the changes, but the findings point to a deliberate shift in the model’s training or fine‑tuning processes. The company’s Claude series has been positioned as a competitor to other large‑language models in the market, and this stylistic refinement could be part of its strategy to differentiate itself.
As AI language models continue to evolve, developers and users alike will be watching how these changes affect the balance between depth of information and readability. For now, Claude Opus 5.5’s new style appears to be a step toward more natural, user‑friendly text generation.
<small>Source: Bleeping Computer — read the original story there.</small>