BenchLeader

LMArena Search Factuality leaderboard

As of 21 Nov 2026, the top score on LMArena Search Factuality is 1248 by GPT-5.6 Sol (OpenAI), ahead of Web Search Agent at 1247, across 43 model configurations with a published result (Arena (LMArena), result dated 19 Nov 2026). Search arena rating with a factuality check folded in at 25% weight: the web-verifiable claims in each search-backed answer are checked, and factual errors count against the model. LMArena Search Factuality is shown for reference only and does not count toward the BenchLeader Index.

Key facts (as of 21 Nov 2026)

  • GPT-5.6 Sol holds the top Arena Search Factuality score at 1248 (Arena (LMArena), 19 Nov 2026).
  • Web Search Agent is second on Arena Search Factuality at 1247 (Arena (LMArena), 19 Nov 2026).
  • Arena Search Factuality: Search arena rating with a factuality check folded in at 25% weight: the web-verifiable claims in each search-backed answer are checked, and factual errors count against the model (Arena (LMArena)).
  • 43 model configurations have a published Arena Search Factuality result on BenchLeader (BenchLeader, 21 Nov 2026).
  • Arena Search Factuality does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).

Search arena rating with a factuality check folded in at 25% weight: the web-verifiable claims in each search-backed answer are checked, and factual errors count against the model.

Published by
Arena (LMArena)
Category
Agents & tools
Index weight
Reference only
Models
43
Data as of
21 Nov 2026

This week on Arena Search Factuality

14 Nov 2026 to 21 Nov 2026, one row per model (its best setting on this board).

CC BY 4.0 — lmarena-ai/leaderboard-dataset on Hugging Face.

43 of 43
#
1GPT-5.6 Solxhigh124863.72026-11-19$8.00
2NewWeb Search Agent1247–2026-11-19–
3GPT-5.5124760.82026-11-19$11.25
4Qwen3.8 Max124260.82026-11-19$3.00
5GPT-6 Astramax123866.82026-11-19$20.00
6GPT-6.1 Solmax123665.82026-11-19$4.00
7Claude Opus 4.6122760.72026-11-19$10.00
8GPT-5.2122455.42026-11-19$4.81
9Grok 4.5122058.92026-11-19$3.00
10ERNIE 5.11214–2026-11-19–
11Claude Opus 5high121464.22026-11-19$10.00
12Claude Fable 5.1high121463.02026-11-19$20.00
13GPT-5.4121257.72026-11-19$5.63
14Claude Fable 5high121263.32026-11-19$20.00
15Claude Opus 4.7121163.02026-11-19$10.00
16Claude Opus 5.5medium120760.52026-11-19$8.00
17GPT-5.1120455.02026-11-19$3.44
18Claude Sonnet 4.6120258.32026-11-19$6.00
19GPT-5.2no reasoning119948.42026-11-19$4.81
20GPT-6 Solmax119862.42026-11-19$4.00
21Gemini 3.1 Pro119859.52026-11-19$4.50
22Grok 4.20 Multi-Agent119858.42026-11-19$1.56
23Grok 4.20 11196–2026-11-19–
24Claude Opus 4.8119558.52026-11-19$10.00
25Gemini 3 Pro119460.32026-11-19$4.50
26NewGPT-6.2 Astramax119368.32026-11-19$16.00
27GPT-6 Lunamax118354.42026-11-19$0.200
28NewGPT-6.2 Solmax118364.42026-11-19$4.00
29Claude Opus 4.5118056.92026-11-19$10.00
30Gemini 3 Flash117958.52026-11-19$1.13
31Claude Sonnet 5117952.52026-11-19$4.00
32Grok 4.3117554.42026-11-19$1.56
33Gemini 2.5 Pro116353.52026-11-19$3.44
34Claude Sonnet 4.5116153.62026-11-19$6.00
35Sonar Reasoning Prohigh1155–2026-11-19$3.50
36Claude Opus 4.1115451.72026-11-19$30.00
37Grok 4.1 Fast1152–2026-11-19$0.275
38Grok 4 Fast1147–2026-11-19$0.275
39GPT-5114456.82026-11-19$3.44
40Claude Opus 4113951.72026-11-19$30.00
41Sonar Prohigh1132–2026-11-19$6.00
42Grok 4113058.02026-11-19$6.00
43o3112960.82026-11-19$3.50

AI agent benchmark leaderboard · All LLM benchmarks

Cite as: BenchLeader, “Arena Search Factuality leaderboard”, data as of 21 Nov 2026, https://www.benchleader.com/benchmarks/lmarena_search_factuality.

Arena Search Factuality: questions

What does Arena Search Factuality measure?
Search arena rating with a factuality check folded in at 25% weight: the web-verifiable claims in each search-backed answer are checked, and factual errors count against the model. Scores are reported in Elo-style ratings from pairwise comparisons; higher is better.
What is the top score on LMArena Search Factuality?
As of 21 Nov 2026, the highest score on Arena Search Factuality is 1248 by GPT-5.6 Sol (result dated 19 Nov 2026), ahead of Web Search Agent at 1247. Scores come from Arena (LMArena) and are re-read daily.
Which AI model leads Arena Search Factuality?
GPT-5.6 Sol leads Arena Search Factuality with 1248 as of 21 Nov 2026 (result dated 19 Nov 2026), ahead of Web Search Agent at 1247.
How many models have Arena Search Factuality results?
43 model configurations have a Arena Search Factuality result on BenchLeader, all taken from Arena (LMArena).
Who runs Arena Search Factuality and how often is it updated?
Arena Search Factuality is published by Arena (LMArena). BenchLeader re-reads the published results once a day and records the date each result was published.
Does Arena Search Factuality count toward the BenchLeader Index?
No. Arena Search Factuality is shown for reference but left out of the composite index.