Search Arena

View rankings across LLMs with web search for real-time information, external knowledge, and grounded citations
Learn more about Search Arena at our blog

Last Updated

Dec 17, 2025

Total Votes

122,219

Total Models

15

/

/

Style Control

	Rank Spread (Upper-Lower)
1	1◄─►3	gemini-3-pro-grounding	1214	±9	4,576	Google	Proprietary
2	1◄─►3	gpt-5.2-search	1211	±12	2,124	OpenAI	Proprietary
3	1◄─►4	gpt-5.1-search	1201	±9	4,458	OpenAI	Proprietary
4	3◄─►4	grok-4-1-fast-search	1185	±10	3,209	xAI	Proprietary
5	5◄─►5	grok-4-fast-search	1168	±6	21,564	xAI	Proprietary
6	6◄─►9	ppl-sonar-reasoning-pro-high	1147	±5	26,092	Perplexity	Proprietary
7	6◄─►10	o3-search	1143	±5	20,917	OpenAI	Proprietary
8	6◄─►11	gemini-2.5-pro-grounding	1142	±5	26,508	Google	Proprietary
9	6◄─►11	grok-4-search	1140	±5	19,228	xAI	Proprietary
10	7◄─►13	claude-opus-4-1-search	1133	±5	26,109	Anthropic	Proprietary
11	8◄─►13	gpt-5-search	1133	±5	20,688	OpenAI	Proprietary
12	10◄─►13	ppl-sonar-pro-high	1129	±5	25,801	Perplexity	Proprietary
13	10◄─►13	claude-opus-4-search	1128	±5	26,255	Anthropic	Proprietary
14	14◄─►15	diffbot-small-xl	1023	±8	6,400	Diffbot	Apache 2.0
15	14◄─►15	api-gpt-4o-search	1011	±10	3,451	OpenAI	Proprietary

Default Leaderboard Plots

Confidence Intervals on Model Strength (via Bootstrapping)

Battle Count for Each Combination of Models (without Ties)

Average Win Rate Against All Other Models (Uniform Sampling and No Ties)

Fraction of Model A Wins for All Non-tied A vs. B Battles