Leaderboard

AI model leaderboard

Which models people prefer when they compare two answers side by side, from LMArena’s open votes. It measures preference, not correctness, and the ± range shows how sure the ranking is. Prices from our API price table.

Agent sessions: task outcome and reliability. Leaderboard of 2026-11-08. Checked .

#ModelScoreVotesLicenceAPI $/M in · out
1Claude Opus 5.5 (Max)
anthropic
0 ±0
2,477,322Proprietary—
2Gemini 4 Carbon (High)
google
0 ±0
8,594,005Proprietary—
3Claude Sonnet 5.5 (Max)
anthropic
0 ±0
1,241,679Proprietary—
4Claude Fable 5.1 (Max)
anthropic
0 ±0
2,440,710Proprietary—
5GPT-6 Astra (Max)
openai
0 ±0
1,318,529Proprietary—
6GPT-6.1 Sol (Max)
openai
0 ±0
3,003,038Proprietary—
7GPT-6.1 Terra (Max)
openai
0 ±0
7,778,687Proprietary—
8Gemini 4 Argon (High)
google
0 ±0
6,590,911Proprietary—
9Muse Wonder Full (Max)
meta
0 ±0
5,895,619Proprietary—
10Claude Fable 5 (High)
anthropic
0 ±0
3,506,790Proprietary—
11Claude Opus 5 (Max)
anthropic
0 ±0
4,646,044Proprietary—
12GPT-6 Sol (Max)
openai
0 ±0
3,219,770Proprietary—
13Kimi K3.1 Pro (Max)
moonshot
0 ±0
6,454,476Kimi K3.1 Pro License—
14Hy4
tencent
0 ±0
2,514,174Proprietary—
15GPT-5.6 Sol (xHigh)
openai
0 ±0
6,593,859Proprietary—
16Grok 4.7 (xHigh)
xai
0 ±0
3,715,248Proprietary—
17Gemini 3.8 Flash (High)
google
0 ±0
8,531,896Proprietary—
18MiMo-V2.6-Pro
xiaomi
0 ±0
5,591,042MIT—
19Muse Spark 1.3 (Max)
meta
0 ±0
8,847,054Proprietary—
20DeepSeek-V4.1-Flash-20261017 (Max)
deepseek
0 ±0
22,460,118MIT—
21Claude Sonnet 5 (High)
anthropic
0 ±0
5,190,733Proprietary—
22Muse Wonder (Max)
meta
0 ±0
2,151,916Proprietary—
23GPT-6 Luna (Max)
openai
0 ±0
41,971,452Proprietary—
24GLM-5.3 (Max)
zai
0 ±0
11,344,109MIT—
25Kimi K3 (Max)
moonshot
0 ±0
16,081,349Kimi K3 license—
26Hy4 preview
tencent
0 ±0
12,008,575Apache 2.0—
27Claude Haiku 5.5 (High)
anthropic
0 ±0
4,305,995Proprietary—
28GLM-5.2 (Max)
zai
0 ±0
9,275,648MIT—
29Qwen3.8-Max
alibaba
0 ±0
8,481,806Proprietary$2 · $6
30MiMo-V2.6-Flash
xiaomi
0 ±0
3,803,116MIT—
31Inkling 2 Small
thinky
0 ±0
529,043Proprietary—
32Grok 4.6 (xHigh)
xai
0 ±0
3,795,827Proprietary—
33GPT-5.6 Terra (xHigh)
openai
0 ±0
3,806,197Proprietary—
34Gemini 3.7 Flash (High)
google
0 ±0
6,197,412Proprietary—
35GPT-6 Luna (Medium)
openai
0 ±0
10,316,806Proprietary—
36MiniMax-M3.1-Flash
minimax
0 ±0
2,724,454MiniMax Community License—
37GLM-5.3-Flash
zai
0 ±0
13,665,236MIT$0.15 · $0.5
38Step 5 Preview
stepfun
0 ±0
2,852,835Proprietary—
39Gemini 3.1 Pro Preview
google
0 ±0
2,960,870Proprietary$2 · $12
40Qwen3.8-27B
alibaba
0 ±0
8,412,983Apache 2.0—
41Qwen3.8-Flash-Next
alibaba
0 ±0
17,026,555qwen-community-1.0—
42Mistral Large 4
mistral
0 ±0
519,381Modified MIT$0.7 · $2.1
43GPT-6 Luna (Low)
openai
0 ±0
3,615,936Proprietary—
44Beam 1.0 501B A23B
reflection
0 ±0
695,650Apache 2.0—
45Qwen3.7-Max
alibaba
0 ±0
2,297,138Proprietary$2.5 · $7.5
46Hy3
tencent
0 ±0
1,233,941Apache 2.0—
47Muse Spark 1.1
meta
0 ±0
8,493,252Proprietary—
48Qwen3.7-Plus
alibaba
0 ±0
1,286,452Proprietary—
49Muse Spark 1.2 (xHigh)
meta
0 ±0
3,866,021Proprietary—
50Spark-X2.6
synlan
0 ±0
1,265,738Proprietary—
51MiniMax-M3
minimax
0 ±0
3,705,686MiniMax Community License—
52Mistral Medium 3.5
mistral
0 ±0
675,056Modified MIT$1.5 · $7.5
53Nemotron 3 Ultra
nvidia
0 ±0
232,397OpenMDW-1.1—
54NVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4)
nvidia
0 ±0
2,205,171OpenMDW-1.1—
55Inkling
thinky
0 ±0
1,739,325Apache 2.0—
56Inkling-Small
thinky
0 ±0
568,000Apache 2.0—
57Solar Pro 4
upstage
0 ±0
559,330Proprietary—

Data: LMArena leaderboard dataset, licensed CC BY 4.0. Prices are matched by model name and may be missing for models without a public API.