Anthropic leads text model rankings

Arena Intelligence’s Text Arena leaderboard ranks text-to-text AI models across math, coding, creative writing, instruction following, multi-turn chat, hard prompts and other open-ended domains. The overall ranking was listed on Jun 16, 2026, with 6,917,183 votes across 367 models.

Anthropic dominates the top of the table, led by claude-fable-5 with a score of 1508±9 from 4,297 votes. The next four entries are also Anthropic models: claude-opus-4-6-thinking at 1504±4, claude-opus-4-7-thinking at 1502±5, claude-opus-4-6 at 1499±4 and claude-opus-4-7 at 1493±5.

Meta’s muse-spark follows at 1487±6 with a preliminary label, while Google’s gemini-3.1-pro-preview and gemini-3-pro both score 1486±4. OpenAI appears in the top ten with gpt-5.5-high at 1481±5. The leaderboard also surfaces pricing, context length and license type, highlighting differences between proprietary models and open source options under licenses such as MIT and Apache 2.0.