Leaderboard

Every model's real record, from actual head-to-head votes. No benchmark, no vendor claim, just what people picked when they saw the answers side by side.

#ModelRatingWin RateBattlesVotesTo First TokenSpeed
-
i
inclusionAI: Ling 3.0 Tiny (free)
Provisional
Compare
1031
Low
100%won 2 of 2
221673 ms449 tok/s
-
D
Dots Studio: Dots3-Note Preview (free)
Provisional
Compare
1016
Low
100%won 1 of 1
112601 ms110.1 tok/s
-
i
inclusionAI: Ling 3.0 Flash Sante (free)
Provisional
Compare
1016
Low
100%won 1 of 1
111550 ms449.9 tok/s
-
N
NVIDIA: Nemotron 3 Super (free)
Provisional
Compare
1016
Low
100%won 1 of 1
1125127 ms432.8 tok/s
-
C
Cohere: North Mini Code (free)
Provisional
Compare
1008
Low
50%won 1 of 2
228672 ms81.4 tok/s
-
M
MiniMax: MiniMax M2.7 (free)
Provisional
Compare
1008
Low
50%won 1 of 2
225466 ms46.9 tok/s
-
D
DeepSeek: DeepSeek V4 Flash 0731 (free)
Provisional
Compare
1000
Low
-won 0 of 0
00273109 ms793.9 tok/s
-
N
NVIDIA: Nemotron 3 Nano 30B A3B (free)
Provisional
Compare
1000
Low
-won 0 of 0
003736 ms222.9 tok/s
-
N
NVIDIA: Nemotron 3.5 Lightning (free)
Provisional
Compare
999
Low
40%won 2 of 5
5574978 ms1059.9 tok/s
-
N
Nex AGI: Nex-N2.5-Pro (free)
Provisional
Compare
994
Low
33%won 1 of 3
3352105 ms187.4 tok/s
-
O
OpenAI: gpt-oss-20b (free)
Provisional
Compare
993
Low
0%won 0 of 1
114544 ms52.8 tok/s
-
F
Free Models Router
Provisional
Compare
992
Low
0%won 0 of 1
113517 ms19.5 tok/s
-
G
Google: Gemma 4 26B A4B (free)
Provisional
Compare
992
Low
0%won 0 of 1
116369 ms56.4 tok/s
-
L
LiquidAI: LFM2.5-2.6B (free)
Provisional
Compare
992
Low
0%won 0 of 1
113102 ms1110.6 tok/s
-
i
inclusionAI: Ling 3.0 Flash Fin (free)
Provisional
Compare
991
Low
0%won 0 of 1
111229 ms552.3 tok/s
-
P
Poolside: Laguna S 2.1 (free)
Provisional
Compare
986
Low
25%won 1 of 4
448181 ms66.2 tok/s
-
N
NVIDIA: Nemotron 3 Ultra (free)
Provisional
Compare
984
Low
0%won 0 of 2
2223579 ms60.6 tok/s
-
P
Poolside: Laguna XS 2.1 (free)
Provisional
Compare
983
Low
0%won 0 of 1
113887 ms196.9 tok/s

Speed is wall clock, request to finish, so a model that buffers its whole answer and a model that streams token by token can sit in the same column honestly. Time to first token is measured separately and shown beside it.