Artificial Intelligence
Claude Opus 5.5 cuts em dashes, lengthens answers
Arena said Anthropic’s Claude Opus 5.5 used far fewer em dashes and shorter sentences than Opus 5, while its average answers also became longer across August and September 2026 samples.
- Arena examined Text Arena responses from August and September 2026.
- It said 10 of 12 writing measures moved in a direction it judged better.
- It found em dashes fell from 15.2 to 0.8 per 1,000 words, while semicolons fell from 6.10 to 1.64.
- It said average sentence length dropped from 12.14 words to 10.03, while average answers rose from 453 words to 481.
Arena examined Text Arena responses from August and September 2026 and said 10 of 12 writing measures moved in a direction it judged better. It said Claude Opus 5.5 showed fewer obvious AI-writing patterns than Opus 5.
Arena said em dashes fell from 15.2 to 0.8 per 1,000 words, which it described as about a 95% drop. It also reported semicolon use falling from 6.10 to 1.64 per 1,000 words.
The comparison also found average sentence length falling from 12.14 words to 10.03. At the same time, average answers rose from 453 words to 481, which made Opus 5.5 the longest-writing Opus model in the sample.
Named in this story
Companies
- Anthropic
- made Claude Opus 5.5
Organisations
- Arena
- analysed the responses and published the comparison
Products and systems
- Text Arena
- supplied the responses Arena analysed
- Claude Opus 5.5
- the newer model in the comparison
- Claude Opus 5
- the earlier model used as the baseline
How the source tells it
The article is upbeat and promotional, relying on benchmark comparisons and the outlet's enthusiasm rather than a plain technical notice.
- vendor boosterism it uses comparative praise and an unstated claim of product improvement
- speculation it extends one benchmark result into a broad prediction about how people will write online