Benchmarks / Arena (formerly LMArena)

Reported by Arena (formerly LMArena)

Arena (formerly LMArena)

Human preference on answers that used live web search.

Last updated 2 Oct 2026

Results dated
24 Aug 2026
Results
34 configurations of 33 models
Unit
Arena rating
Licence
Creative Commons Attribution 4.0 International

Arena (formerly LMArena) in brief

What it measures. Arena (formerly LMArena) is a third-party benchmark, shown here with credit. Measured here: Human preference on answers that used live web search.

Latest results. , 34 configurations ranked.

Top 3 on Overall (Arena rating, higher is better):

  1. GPT-5.6 Sol (extra-high reasoning) (OpenAI): 1,256
  2. Claude Fable 5 (Anthropic): 1,229
  3. GPT-5.5 (OpenAI): 1,224

How to cite. Arena (formerly LMArena), via Spring Prompt, results dated 24 August 2026. https://springprompt.com/benchmarks/lmarena These results belong to Arena (formerly LMArena), not to us. Licence: Creative Commons Attribution 4.0 International. Credit Arena (formerly LMArena) first.

Overall: Claude Sonnet 4.5

Top 15 of 34 results · Arena rating, higher is better · lines show the 95% range. Choose a model to highlight it.Clear highlight

  1. 1 GPT-5.6 Sol (extra-high reasoning)OpenAI 1,256
  2. 2 Claude Fable 5Anthropic 1,229
  3. 3 GPT-5.5OpenAI 1,224
  4. 4 Claude Opus 4.6Anthropic 1,223
  5. 5 Claude Opus 4.7Anthropic 1,212
  6. 6 Gemini 3.1 Pro PreviewGoogle 1,208
  7. 7 Grok 4.20 Multi-Agent Beta (0309)xAI 1,205
  8. 8 Grok 4.5xAI 1,202
  9. 9 Gemini 3 Pro PreviewGoogle 1,201
  10. 9 Claude Sonnet 4.6Anthropic 1,201
  11. 11 Grok 4.20 Beta 1xAI 1,197
  12. 12 Claude Sonnet 5Anthropic 1,196
  13. 13 Claude Opus 4.8Anthropic 1,195
  14. 13 GPT-5.4OpenAI 1,195
  15. 15 Grok 4.1 FastxAI 1,194
  16. 23 Claude Sonnet 4.5Anthropic 1,175

Full results

Arena Search: overall, Arena rating, higher is better
#ModelOverall · 95% range
Arena rating, higher is better
Price
$ per million tokens, in / out
1 GPT-5.6 Sol (extra-high reasoning)OpenAI
1,256
1,248–1,263
$4 / $20
2 Claude Fable 5Anthropic
1,229
1,221–1,237
$10 / $50
3 GPT-5.5OpenAI
1,224
1,218–1,229
$5 / $30
4 Claude Opus 4.6Anthropic
1,223
1,218–1,228
$5 / $25
5 Claude Opus 4.7Anthropic
1,212
1,207–1,218
$5 / $25
6 Gemini 3.1 Pro PreviewGoogle
1,208
1,202–1,213
$2 / $12
7 Grok 4.20 Multi-Agent Beta (0309)xAI
1,205
1,199–1,210
–
8 Grok 4.5xAI
1,202
1,195–1,209
$2 / $6
9 Gemini 3 Pro PreviewGoogle
1,201
1,196–1,207
–
9 Claude Sonnet 4.6Anthropic
1,201
1,196–1,206
$3 / $15
11 Grok 4.20 Beta 1xAI
1,197
1,191–1,203
–
12 Claude Sonnet 5Anthropic
1,196
1,189–1,203
$2 / $10
13 Claude Opus 4.8Anthropic
1,195
1,189–1,201
$5 / $25
13 GPT-5.4OpenAI
1,195
1,189–1,200
$2.50 / $15
15 Grok 4.1 FastxAI
1,194
1,189–1,199
–
16 ERNIE 5.1Baidu
1,193
1,184–1,203
–
16 Grok 4.3xAI
1,193
1,188–1,199
$1.25 / $2.50
18 o3OpenAI
1,192
1,185–1,199
$2 / $8
19 Gemini 3 Flash PreviewGoogle
1,191
1,187–1,196
$0.50 / $3
20 Claude Opus 4.5Anthropic
1,187
1,181–1,193
$5 / $25
21 GPT-5.1OpenAI
1,183
1,178–1,189
$1.25 / $10
21 GPT-5OpenAI
1,183
1,175–1,190
$1.25 / $10
23 Claude Sonnet 4.5Anthropic
1,175
1,170–1,180
$3 / $15
24 Claude Opus 4.1Anthropic
1,167
1,162–1,172
$15 / $75
25 GPT-5.2OpenAI
1,162
1,156–1,167
$1.75 / $14
26 Grok 4 FastxAI
1,160
1,155–1,165
–
27 Claude Opus 4Anthropic
1,151
1,145–1,156
–
28 Sonar Pro (high search context)Perplexity
1,143
1,137–1,149
–
29 Gemini 2.5 ProGoogle
1,142
1,137–1,147
$1.25 / $10
30 Sonar Reasoning Pro (high search context)Perplexity
1,138
1,132–1,144
–
31 GPT-5.2 (no reasoning)OpenAI
1,137
1,131–1,143
$1.75 / $14
32 Grok 4xAI
1,121
1,115–1,128
–
33 GPT-4o Search (API)OpenAI
1,081
1,070–1,092
–
34 Diffbot Small XLDiffbot
1,062
1,053–1,071
–

Swipe the table sideways for more columns.

Source: Arena (formerly LMArena), via Spring Prompt, springprompt.com/benchmarks/lmarena. Test your own prompt on these models →

Ranks follow the score as shown, so equal numbers share a rank. Where two models' ranges overlap, the gap between them may not be real. Results as published by Arena (formerly LMArena); we do not re-run them.

Progress over time

Overall, plotted at each model's release date, higher is better. Each model counts once, at its best setting. Scores are today's, not launch day's; 8 models without a known release date are left off. Hover or tap a mark for its model.

Best score so far

Every model at its release date. The line moves each time a new release beats every earlier one, and each new best is named where there is room.

1,1001,1501,2001,2501,300 Apr 2025JulOct2026AprJulOct Gemini 2.5 Proo3Gemini 3 Pro PreviewClaude Opus 4.6GPT-5.5Claude Fable 5GPT-5.6 Sol

Best score by lab

Each lab's best so far, stepping whenever one of its releases beats its own earlier models.

OpenAIAnthropicGooglexAI

1,1001,1501,2001,2501,300 Apr 2025JulOct2026AprJulOct OpenAIAnthropicGooglexAI

What it measures

Human preference on answers that used live web search.

What it does not measure

Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.

Contains data from the Arena Leaderboard Dataset by Arena, licensed under CC BY 4.0. Licence: Creative Commons Attribution 4.0 International.