Rankings

Arena.ai leaderboards: Elo op basis van blind head-to-head stemmen. Kies een categorie.

Document

PDF- en documentbegrip.

14 september 2026, 07:09

44 modellen

#	Model	Elo	±	Stemmen
	claude-opus-5-high Anthropic · Proprietary ±7 · 8.8k stemmen	1516	±7	8.8k
	claude-fable-5.1-max Anthropic · Proprietary ±15 · 1.4k stemmen	1513	±15	1.4k
	claude-opus-4-6-high Anthropic · Proprietary ±6 · 27.9k stemmen	1507	±6	27.9k
4	claude-opus-4-6 Anthropic · Proprietary ±6 · 41.3k stemmen	1507	±6	41.3k
5	claude-fable-5 Anthropic · Proprietary ±8 · 8.5k stemmen	1496	±8	8.5k
6	claude-opus-4-7 Anthropic · Proprietary ±6 · 22.2k stemmen	1495	±6	22.2k
7	claude-opus-4-7-high Anthropic · Proprietary ±7 · 22.0k stemmen	1495	±7	22.0k
8	gpt-5.5 OpenAI · Proprietary ±6 · 21.3k stemmen	1486	±6	21.3k
9	gpt-5.5-high OpenAI · Proprietary ±6 · 20.9k stemmen	1484	±6	20.9k
10	gpt-5.6-sol-xhigh OpenAI · Proprietary ±9 · 4.8k stemmen	1483	±9	4.8k
11	claude-sonnet-4-6 Anthropic · Proprietary ±6 · 57.8k stemmen	1482	±6	57.8k
12	claude-opus-4-8-high Anthropic · Proprietary ±7 · 11.6k stemmen	1475	±7	11.6k
13	gpt-5.6-terra-xhigh OpenAI · Proprietary ±8 · 5.8k stemmen	1472	±8	5.8k
14	gpt-5.4 OpenAI · Proprietary ±6 · 33.3k stemmen	1471	±6	33.3k
15	muse-spark-1.3-max Meta · Proprietary ±18 · 1.0k stemmen	1471	±18	1.0k

Elo-score op basis van blind head-to-head stemmen. Hoger is beter. ± is het 95% betrouwbaarheidsinterval. Zelfde formaat als Arena.ai.