Claude Opus 5.5 ranks first in Text Arena: what the official data says
Arena confirms 1,509 points for Opus 5.5 (High). We examine the ranking, its uncertainty and Anthropic’s launch.
Checked against the official sources linked below.
Confirmed on Arena’s leaderboard
The official Text Arena table checked on September 27, 2026 ranks claude-opus-5.5-high first at 1,509 ±12 points with 2,307 votes. Opus 4.6 (High) follows at 1,505 ±3. The table is dated September 25: these numbers describe that snapshot, not a permanent position. [1]
First place and uncertainty
Arena separates score order from the rank spread implied by confidence intervals. Opus 5.5 has a spread of 1 to 10 in this snapshot. The highest score does not establish a definitive statistical advantage over every other model. Both fields matter when interpreting the headline. [1] [2]
What Anthropic announced
Anthropic introduced Opus 5.5 on September 22. Base API pricing is $4 per million input tokens and $20 per million output tokens. The vendor reports lower running costs and faster output than Opus 5 in its evaluations. These are Anthropic’s results, not tests conducted for this article. [3]
What it contributes to a development decision
My editorial interpretation: the ranking helps shortlist a candidate. For a project, I would compare accepted changes, corrections and total review cost. A general first place cannot replace testing the routes, integrations and constraints of the actual repository.
Before adopting a model, record:
- Model version and effort settings.
- Identical tasks and acceptance criteria for each candidate.
- Introduced errors, retries and human review time.