Showdown Leaderboard - LLMs
Real people. Real conversations. Real rankings.
Showdown ranks AI models based on how they perform in real-world use— not synthetic tests or lab settings. Votes are blind, optional, and organic, so rankings reflect authentic preferences.Methodology & Technical Report→0 promptsReal conversation prompts compared across models through pairwise votes.
0 usersFrom 80+ countries and 70+ languages, spanning all backgrounds and professions.
SEAL Leaderboard - LLMs
RANK ↑
MODEL ↑↓
VOTES ↑↓
SCORE ↑↓
1
gpt-5-chat
gpt-5-chat13532
1096.41
-5.89 +4.29
1
claude-sonnet-4-5-20250929
claude-sonnet-4-5-202509297625
1091.76
-4.45 +4.29
2
qwen3-235b-a22b-2507-v1
qwen3-235b-a22b-2507-v14025
1084.08
-6.72 +6.08
3
claude-opus-4-1-20250805
claude-opus-4-1-2025080516479
1083.07
-3.98 +3.11
5
claude-sonnet-4-20250514
claude-sonnet-4-2025051418268
1070.55
-2.48 +2.99
5
claude-sonnet-4-5-20250929 (Thinking)
claude-sonnet-4-5-20250929 (Thinking)7513
1068.57
-4.81 +5
5
claude-opus-4-20250514
claude-opus-4-2025051416147
1064.92
-4.2 +4.52
5
claude-haiku-4-5-20251001
claude-haiku-4-5-202510015063
1062.16
-6.62 +7.63
7
gpt-4.1-2025-04-14
gpt-4.1-2025-04-1417661
1058.75
-3.68 +2.87
8
claude-opus-4-1-20250805 (Thinking)
claude-opus-4-1-20250805 (Thinking)15196
1053.63
-4.22 +4.55
10
gemini-2.5-pro-preview-06-05
gemini-2.5-pro-preview-06-0515357
1045.74
-4.31 +4.15
11
claude-opus-4-20250514 (Thinking)
claude-opus-4-20250514 (Thinking)15639
1038.50
-4.01 +3.56
12
claude-sonnet-4-20250514 (Thinking)
claude-sonnet-4-20250514 (Thinking)18017
1036.65
-4.01 +3.23
12
claude-haiku-4-5-20251001 (Thinking)
claude-haiku-4-5-20251001 (Thinking)4861
1029.13
-6.13 +8.37
14
gemini-2.5-flash-preview-05-20
gemini-2.5-flash-preview-05-2018363
1019.72
-2.73 +4.05
15
20435
1019.94
-3.58 +2.9
17
llama4-maverick-instruct-basic
llama4-maverick-instruct-basic18919
1000.00
-3.46 +4.67
18
19835
988.83
-3.28 +3.42
19
deepseek-r1-0528
deepseek-r1-05284275
967.13
-8.72 +7.4
20
14014
951.83
-4.66 +5.18
* This model’s API does not consistently return Markdown-formatted responses. Since raw outputs are used in head-to-head comparisons, this may affect its ranking.
