Benchmarks / Composite

Arena Text

Crowdsourced pairwise human preference on open-ended text prompts (style control on); Bradley-Terry rating.

212models
–Open Index weight
–last updated
LMArenapublisher

Arena Text

Higher is better · best effort per model

Chart settings
212 of 212 selected

full table below

#ModelEffortScoreCost / taskHarness
1Gemini 4high1524––
2Opus 5.5high1508––
3Fable 5high1504––
4Opus 4.6high1503––
5Fable 5.1max1502––
6Opus 4.7high1501––
7Gemini 3.8 Flashhigh1496––
8Muse Spark 1.3max1494––
9Muse Spark 1.2xhigh1493––
10Muse Spark 1.1max1490––
11Opus 5max1490––
12Muse Spark–1489––
13Gemini 3.7 Flashhigh1488––
14Kimi K3max1487––
15Gemini 3.1 Prohigh1487––
16Gemini 3 Prohigh1485––
17GPT-5.6 Solxhigh1485––
18GPT-6.1 Solmax1484––
19Gemini 3.6 Flashhigh1483––
20Qwen 3.8 Maxmax1482––
21Opus 4.8high1481––
22GPT-5.5high1481––
23MiMo V2.6 Prohigh1479––
24GLM 5.3max1479––
25Gemini 3.5 Flashhigh1477––
26GPT-6 Astramax1477––
27GLM 5.2max1476––
28GPT-5.2xhigh1476––
29GPT-5.4high1475––
30Sonnet 5.5xhigh1475––
31Grok 4.20xhigh1475––
32DeepSeek V4.1 Flashmax1475––
33Qwen 3.7 Maxmax1474––
34GLM 5.3 Flashmax1474––
35Opus 4.5high1474––
36GPT-5.5 Instant–1473––
37Gemini 3 Flashhigh1473––
38Sonnet 4.6max1472––
39MiMo V2.5 Proxhigh1467––
40GPT-5.6 Terraxhigh1466––
41Grok 4.5max1466––
42Grok 4.1–1465––
43GLM 5.1max1465––
44DeepSeek V4 Promax1465––
45Qwen 3.5 Max–1464––
46Sonnet 5high1461––
47Kimi K2.6max1461––
48Qwen 3.6 Maxmax1461––
49GPT-6 Solmax1459––
50GLM 5high1457––
51Sonnet 4.5high1456––
52GPT-5.1high1456––
53Qwen 3.7 Plushigh1456––
54Gemini 3.5 Flash-Litexhigh1455––
55Grok 4.6high1454––
56Hy3xhigh1453––
57GPT-5.6 Lunaxhigh1453––
58Gemma 4 31Bmax1452––
59MiMo V2.6 Flashhigh1451––
60GPT-5.3xhigh1450––
61Kimi K2.5max1450––
62Opus 4.1high1450––
63MiMo V2 Pro–1448––
64GPT-5.4 Minihigh1447––
65Gemini 2.5 Prohigh1445––
66GPT-4.5none1445––
67Grok 4.7xhigh1444––
68Qwen 3.6 Plusmax1443––
69GPT-6 Lunamax1443––
70Grok 4.3xhigh1442––
71Qwen 3.5 397B-A17Bmax1442––
72GLM 4.7none1441––
73MiniMax M3max1441––
74Inklingmax1441––
75DeepSeek V4 Flashmax1439––
76Qwen 3.8 27Bmax1438––
77Gemma 4 26Bxhigh1438––
78GPT-5high1435––
79Qwen 3 Maxmax1434––
80MiMo V2.5xhigh1434––
81GLM 5 Vhigh1434––
82Gemini 3.1 Flash-Litehigh1433––
83Mistral Large 4high1432––
84o3high1432––
85MiMo V2–1431––
86Grok 4.1 Fasthigh1430––
87Kimi K2medium1430––
88Mistral Medium 3.5high1426––
89Opus 4medium1426––
90Nemotron 3 Ultramax1426––
91DeepSeek V3.2high1425––
92GLM 4.6none1424––
93Qwen 3 235B-A22Bxhigh1422––
94DeepSeek R1.0528–1421––
95Grok 4 Fastnone1419––
96Kimi K2.0905–1419––
97Kimi K2.0711–1418––
98DeepSeek V3.1none1417––
99Qwen 3.5 122B-A10Bnone1416––
100MiniMax M2.7high1415––
101GPT-4.1high1415––
102Haiku 4.5xhigh1414––
103Mistral Large 3none1414––
104Qwen 3high1413––
105Grok 3none1411––
106Grok 4.0709–1411––
107GLM 4.5xhigh1411––
108Gemini 2.5 Flashmedium1409––
109Qwen 3.5 27Bhigh1409––
110Mistral Medium 2508–1408––
111Inkling Smallxhigh1404––
112o1high1402––
113Sonnet 4high1402––
114GPT-5.4 Nanohigh1401––
115DeepSeek R1xhigh1398––
116Qwen 3.5 Flashmax1396––
117DeepSeek V3.0324–1395––
118Qwen 3.5 35B-A3Blow1394––
119MiMo V2 Flashhigh1392––
120o4 minihigh1391––
121MiniMax M2.5high1391––
122GPT-5 Minihigh1390––
123Sonnet 3.7medium1388––
124Qwen 3 Codermedium1388––
125Mistral Medium 2505–1387––
126MiniMax M2.1–1384––
127GPT-4.1 Minihigh1383––
128Qwen 3 30B-A3Bnone1382––
129Gemini 2.5 Flash Litenone1379––
130GLM 4.6 Vlow1378––
131Sonnet 3.5medium1374––
132Qwen 2.5 Maxnone1374––
133GLM 4.5 Air–1373––
134Gemma 3 27B ITmedium1365––
135GLM 4.7 Flashnone1365––
136MiniMax M1–1364––
137Grok 3 Minihigh1364––
138o3 minihigh1364––
139Gemini 2.0 Flashmedium1360––
140DeepSeek V3none1358––
141Mistral Small 2506–1356––
142Gemini 2.0 Flash-Litenone1354––
143GLM 4.5 Vnone1352––
144gpt-oss-120bhigh1352––
145Nemotron 3.5 Lightning–1351––
146Gemini 1.5 Pronone1351––
147Llama 3.1max1348––
148Qwen 3 32Bnone1347––
149GPT-4ohigh1346––
150MiniMax M2–1343––
151Mercury 2high1343––
152Llama 3.3none1343––
153GLM 4–1343––
154Gemma 3 12B IT–1342––
155GPT-5 Nanohigh1338––
156o1 minihigh1337––
157Grok 2.2024–1336––
158Llama 3.1 405bnone1335––
159Llama 4 Maverickmedium1327––
160Haiku 3.5medium1325––
161GPT-4high1325––
162DeepSeek V2.5none1323––
163Opus 3none1322––
164GPT-4.1 Nanohigh1322––
165Llama 4 Scoutmedium1322––
166Llama 3.3 70bnone1318––
167GPT-4o minihigh1318––
168gpt-oss-20bhigh1317––
169Qwen 2.5 Plus–1315––
170Mistral Large 2407–1314––
171GPT-4.0125–1313––
172GPT-4.1106–1313––
173Gemini 1.5 Flashnone1309––
174Grok 2 Mini–1308––
175Mistral Large 2411–1306––
176Magistral Medium 2506–1305––
177Gemma 3 4B IT–1303––
178Qwen 2.5 72Bnone1303––
179Mistral Small 3.1none1303––
180Llama 3.1 70bnone1293––
181Gemma 2 27B IT–1290––
182GPT-4.0314–1288––
183Sonnet 3none1281––
184Gemma 2 9B IT–1280––
185Llama 3 70bnone1277––
186GPT-4.0613–1276––
187Mistral Smallnone1274––
188GLM 4.0520–1273––
189Qwen 2.5 Codernone1271––
190Qwen 2 72Bnone1262––
191Haiku 3medium1262––
192Mistral Large 2402–1243––
193Qwen 1.5 110Bnone1234––
194Qwen 1.5 72B–1234––
195GPT-3.5high1226––
196Llama 3 8bnone1224––
197Mistral Mediumnone1223––
198Llama 3.1 8bmedium1211––
199Qwen 1.5 32B–1204––
200Gemma 2 2B IT–1200––
201Qwen 1.5 14B–1191––
202Gemma 1.1 7B IT–1183––
203Mistralnone1176––
204Llama 2 70bnone1171––
205Llama 3.2 3bnone1167––
206Qwen 1.5 7B–1144––
207Llama 2 13bnone1142––
208Gemma 1.1 2B IT–1117––
209Llama 3.2 1bnone1111––
210Llama 2 7bnone1108––
211Qwen 1.5 4B–1091––
212Llama 1 3b–975––