Benchmark Almanac

Arena Search Factuality

gpt-5.2-search 1,273 tops its leaderboard Grok 4.20 1,218 leads the models we track
Not in model cards

Arena combines user preferences with independent checks of verifiable claims in model responses.

Arena publishes this ranking separately from the unadjusted human preference ranking.

Made by
Arena
Released
July 2026
Kind
Long-form factuality
Level
Everyday
Scored
Graded by a model
Answers
Open-ended
Measures
Propensity
Behavior
Hallucination
In and out
Text → Text
Interaction
Single turn
Environment
Browser
Items from
Live play or votes
Refresh
Live
Run by
A host
In views
Kept fresh
Strongest model tested
Claude Fable 5 #12 on the capability index
Last model added
June 2026
Scores refreshed
November 21st, 2026
32
Entries on its own leaderboard, 7 models we track

gpt-5.2-search tops the leaderboard with 1,273 Grok 4.20 leads with 1,218 among models released by June 2026

Best score per entry from the benchmark's own leaderboard, read on November 21st, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is Grok 4.20, at 1,218. This view shows only models we track, each at its best.

The newest model with a score here came out in June 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

Bars start at 1,050.

  1. 1 gpt-5.2-search 1,273
  2. 2 gpt-5.2-search-non-reasoning 1,272
  3. 3 gpt-5.5-search 1,269
  4. 4 gpt-5.4-search 1,263
  5. 5 gpt-5.6-sol-search-xhigh 1,241
  6. 6 grok-4.5-search 1,227
  7. 7 gpt-5.1-search 1,226
  8. 8 Grok 4.20 1,218
  9. 9 gemini-2.5-pro-grounding 1,217
  10. 10 gpt-5-search 1,214
  11. 11 Grok 4.3 1,211
  12. 12 Grok 4.20 Multi-agent 1,203
  13. 13 Claude Opus 4.8 1,196
  14. 14 claude-opus-4-search 1,195
  15. 15 claude-opus-4-5-search 1,191
  16. 16 claude-sonnet-5-search 1,190
  17. 17 Claude Fable 5 1,185
  18. 18 claude-opus-4-6-search 1,184
  19. 19 ppl-sonar-pro-high 1,183
  20. 20 gemini-3.1-pro-grounding 1,182
  21. 21 claude-sonnet-4-5-search 1,182
  22. 22 gemini-3-pro-grounding 1,181
  23. 23 claude-opus-4-1-search 1,178
  24. 24 ppl-sonar-reasoning-pro-high 1,172
  25. 25 Claude Opus 4.7 1,171
  26. 26 claude-sonnet-4-6-search 1,169
  27. 27 gemini-3-flash-grounding 1,153
  28. 28 grok-4-search 1,128
  29. 29 grok-4-1-fast-search 1,123
  30. 30 grok-4-fast-search 1,112
  31. 31 ERNIE 5.1 1,097
  32. 32 o3-search 1,082

Source: Official Arena Search Factuality leaderboard

#EntryScore ±SettingDateSource
1 gpt-5.2-search 1,273 ±7.4 Official leaderboard
2 gpt-5.2-search-non-reasoning 1,272 ±7.6 Official leaderboard
3 gpt-5.5-search 1,269 ±8.5 Official leaderboard
4 gpt-5.4-search 1,263 ±7.1 Official leaderboard
5 gpt-5.6-sol-search-xhigh 1,241 ±18.5 Official leaderboard
6 grok-4.5-search 1,227 ±19.5 Official leaderboard
7 gpt-5.1-search 1,226 ±7.9 Official leaderboard
8 Grok 4.20 1,218 ±8.2 Feb 2026 Official leaderboard
9 gemini-2.5-pro-grounding 1,217 ±7.6 Official leaderboard
10 gpt-5-search 1,214 ±26.1 Official leaderboard
11 Grok 4.3 1,211 ±8.7 Apr 2026 Official leaderboard
12 Grok 4.20 Multi-agent 1,203 ±7.4 Mar 2026 Official leaderboard
13 Claude Opus 4.8 1,196 ±9 May 2026 Official leaderboard
14 claude-opus-4-search 1,195 ±14.8 Official leaderboard
15 claude-opus-4-5-search 1,191 ±8.3 Official leaderboard
16 claude-sonnet-5-search 1,190 ±13.4 Official leaderboard
17 Claude Fable 5 1,185 ±15 Jun 2026 Official leaderboard
18 claude-opus-4-6-search 1,184 ±6.8 Official leaderboard
19 ppl-sonar-pro-high 1,183 ±15.3 Official leaderboard
20 gemini-3.1-pro-grounding 1,182 ±7.6 Official leaderboard
21 claude-sonnet-4-5-search 1,182 ±7.4 Official leaderboard
22 gemini-3-pro-grounding 1,181 ±10.3 Official leaderboard
23 claude-opus-4-1-search 1,178 ±8.2 Official leaderboard
24 ppl-sonar-reasoning-pro-high 1,172 ±14.6 Official leaderboard
25 Claude Opus 4.7 1,171 ±8.8 Apr 2026 Official leaderboard
26 claude-sonnet-4-6-search 1,169 ±7 Official leaderboard
27 gemini-3-flash-grounding 1,153 ±7.3 Official leaderboard
28 grok-4-search 1,128 ±31.3 Official leaderboard
29 grok-4-1-fast-search 1,123 ±8.1 Official leaderboard
30 grok-4-fast-search 1,112 ±11.3 Official leaderboard
31 ERNIE 5.1 1,097 ±30.9 May 2026 Official leaderboard
32 o3-search 1,082 ±32.6 Official leaderboard
#ModelScore ±SettingDateSource
1 Grok 4.20 1,218 ±8.2 Feb 2026 Official leaderboard
2 Grok 4.3 1,211 ±8.7 Apr 2026 Official leaderboard
3 Grok 4.20 Multi-agent 1,203 ±7.4 Mar 2026 Official leaderboard
4 Claude Opus 4.8 1,196 ±9 May 2026 Official leaderboard
5 Claude Fable 5 1,185 ±15 Jun 2026 Official leaderboard
6 Claude Opus 4.7 1,171 ±8.8 Apr 2026 Official leaderboard
7 ERNIE 5.1 1,097 ±30.9 May 2026 Official leaderboard

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

xAI
Own leaderboard Not tested since Jul 2026
Grok 4.20 1,218

Source: Official Arena Search Factuality leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard
1,100 1,150 1,200 2026 1,100 1,150 1,200 2026

Source: Official Arena Search Factuality leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Factuality

All 164 in Factuality
  • SimpleVQA In use Models answer 2,025 short factual questions about images across 9 tasks and 9 topics. An LLM judge grades answers against fixed references. Qwen3.7-Plus 81.7%
  • HallusionBench In use Models answer 1,129 yes or no questions about 346 images, built in pairs to expose language hallucination and visual illusion. Scored by accuracy. Keye-VL-2.0-30B-A3B 73.5%
  • SimpleQA Verified In use 1,000 human-verified, de-duplicated short-answer factual questions measuring parametric knowledge. Gemini 3.1 Pro 77.5%
  • AA-Omniscience v1.1 In use Models answer 3,000 knowledge questions in six domains without tools and may abstain. The index adds correct answers and subtracts wrong ones. Inkling 2 Small 16.1
  • GaRAGe In use Models write long-form RAG answers to 2,366 questions with human-annotated grounding passages. Scoring checks whether answers use the relevant passages. Amazon Nova Pro 60.7%
  • Kimi Hallu Bench In use Models answer hard prompts drawn from practical work tasks. Fact-checking agents score the share of claims that are hallucinated. GPT-6 Astra 4.20%

Data audit

8 of 9 checks pass, 1 warning. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark

Catalog record 1 warning 1 pass
  • Warning Task access We don’t know whether the tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (32 entries).
  • Pass Models scored 7 models we track, plus 25 other entries on its leaderboard.
  • Pass Newest score The newest is from June 2026, for Claude Fable 5.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Claude Fable 5 (#12).
  • Note Headroom Scored as a rating, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on arena.ai: 7 rows for models we track and 25 other entries at the last read.
  • Pass Last read November 21st, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check