1,503
Text Arena rating of claude-opus-4-6 in early 2026, tied at the top with its thinking variant; gemini-3.1-pro-preview 1,500

What the chart shows

The Arena ranks models by users’ blind votes between two answers to the same prompt. In early 2026 claude-opus-4-6 and claude-opus-4-6-thinking tied at the top of the text leaderboard with 1,503, followed by gemini-3.1-pro-preview (1,500) and grok-4.20-beta1 (1,495).

The top 15 span 46 points, from 1,503 down to gpt-5.1-high’s 1,457. The ratings’ 95% confidence intervals range from ±4 to ±10 points and overlap for many neighbouring models, so small differences in rank are not meaningful.

Key numbers

MeasureValue
claude-opus-4-6 (95% CI 1,496–1,510)1,503
claude-opus-4-6-thinking (95% CI 1,495–1,511)1,503
gemini-3.1-pro-preview1,500
grok-4.20-beta11,495
gpt-5.1-high (15th)1,457

How to read it

  • Ratings from the Arena text leaderboard (arena.ai) in early 2026; top 15 models only.
  • Ratings reflect user preference, which can favour longer or better-styled answers over correct ones.
  • The report gives the leader’s rating as about 1,510; the data shows 1,503 for both Claude Opus 4.6 entries, which we use.
  • Elo-like ratings have no true zero, so the chart shows dots on an axis starting at 1,440: compare the differences between models.
Source: Arena, 2026, via the AI Index Report 2026 (Stanford HAI), figure 2.2.2: Text Arena: Elo rating. Chart and text by AI Adoption.
Numbers checked against the published data on 20 October 2026.
Download the data (CSV) · Cite this chart