lmarena text Benchmark: Scores and Sources

Published result; benchmark version and evaluation conditions remain in the id and result note.

pillar: preference · weight 1 within pillar · unit: elo (higher is better) · official board ↗

Row	Model	Result	Normalized (0–100)	Confidence
1	Claude Opus 5.5 Anthropic	1511.7 elo	82.6	100% confidence 100 percent, Full
2	Claude Fable 5.1 Anthropic	1510.6 elo	82.5	100% confidence 100 percent, Full
3	Claude Opus 5 Anthropic	1502.1 elo	81.9	100% confidence 100 percent, Full
4	Claude Opus 4.6 Anthropic	1497.6 elo	81.6	100% confidence 100 percent, Full
5	Gemini 3.8 Flash Google	1497.0 elo	81.5	100% confidence 100 percent, Full
6	Claude Fable 5 Anthropic	1491.7 elo	81.1	100% confidence 100 percent, Full
7	MiMo-V2.6-Pro xiaomi	1491.0 elo	81.1	88% confidence 88 percent, High
8	Muse Spark 1.3 Meta	1489.9 elo	81.0	93% confidence 93 percent, High
9	Gemini 3.7 Flash Google	1487.0 elo	80.8	100% confidence 100 percent, Full
10	Claude Opus 4.7 Anthropic	1483.3 elo	80.5	100% confidence 100 percent, Full
11	Muse Spark 1.2 Meta	1483.2 elo	80.5	53% confidence 53 percent, Medium
12	Qwen3.8 Max Alibaba / Qwen	1481.8 elo	80.4	85% confidence 85 percent, High
13	Gemini 3.1 Pro Preview Google	1480.2 elo	80.2	100% confidence 100 percent, Full
14	Gemini 3.6 Flash Google	1479.5 elo	80.2	100% confidence 100 percent, Full
15	Muse Spark 1.1 Meta	1479.2 elo	80.2	61% confidence 61 percent, Medium
16	Gemini 3.5 Flash Google	1477.7 elo	80.0	100% confidence 100 percent, Full
17	Kimi K3 Moonshot AI	1475.9 elo	79.9	100% confidence 100 percent, Full
18	GLM-5.3 Z.ai	1471.4 elo	79.5	93% confidence 93 percent, High
19	GLM-5.2 Z.ai	1470.4 elo	79.4	100% confidence 100 percent, Full
20	GLM-5.3-Flash Z.ai	1469.6 elo	79.4	100% confidence 100 percent, Full
21	GPT-5.5 OpenAI	1466.8 elo	79.1	100% confidence 100 percent, Full
22	Claude Sonnet 5.5 Anthropic	1466.6 elo	79.1	93% confidence 93 percent, High
23	MiMo-V2.5-Pro xiaomi	1465.0 elo	79.0	88% confidence 88 percent, High
24	DeepSeek V4.1 Flash DeepSeek	1462.5 elo	78.8	69% confidence 69 percent, Medium
25	GLM-5.1 Z.ai	1461.5 elo	78.7	64% confidence 64 percent, Medium
26	Gemini 2.5 Pro Google	1457.8 elo	78.4	90% confidence 90 percent, High
27	Claude Sonnet 4.6 Anthropic	1457.8 elo	78.4	100% confidence 100 percent, Full
28	MiMo-V2.6-Flash xiaomi	1456.4 elo	78.3	88% confidence 88 percent, High
29	GPT-5.6 Sol OpenAI	1456.3 elo	78.3	100% confidence 100 percent, Full
30	Kimi K2.6 Moonshot AI	1455.5 elo	78.2	85% confidence 85 percent, High
31	Qwen3.7 Plus Alibaba / Qwen	1454.6 elo	78.1	64% confidence 64 percent, Medium
32	Claude Opus 4.8 Anthropic	1454.1 elo	78.1	100% confidence 100 percent, Full
33	GPT-5.4 OpenAI	1452.2 elo	77.9	100% confidence 100 percent, Full
34	DeepSeek V4 Pro DeepSeek	1451.0 elo	77.8	85% confidence 85 percent, High
35	Grok 4.20 (Reasoning) xAI	1450.8 elo	77.8	80% confidence 80 percent, High
36	Grok 4.5 xAI	1448.1 elo	77.6	100% confidence 100 percent, Full
37	Claude Opus 4.5 Anthropic	1448.1 elo	77.6	100% confidence 100 percent, Full
38	Step 5 Preview stepfun	1447.6 elo	77.5	75% confidence 75 percent, Medium
39	Qwen3.6 Max Preview Alibaba / Qwen	1446.8 elo	77.5	64% confidence 64 percent, Medium
40	GPT-5.6 Terra OpenAI	1446.3 elo	77.4	100% confidence 100 percent, Full
41	GLM-5 Z.ai	1446.3 elo	77.4	90% confidence 90 percent, High
42	GPT-6.1 Sol OpenAI	1445.6 elo	77.3	100% confidence 100 percent, Full
43	Kimi K2.5 Moonshot AI	1445.2 elo	77.3	100% confidence 100 percent, Full
44	Gemma 4 31B IT Google	1443.4 elo	77.2	64% confidence 64 percent, Medium
45	Claude Sonnet 5 Anthropic	1442.9 elo	77.1	93% confidence 93 percent, High
46	GPT-6 Astra OpenAI	1442.3 elo	77.1	100% confidence 100 percent, Full
47	Gemini 3 Flash Preview Google	1442.3 elo	77.1	93% confidence 93 percent, High
48	Inkling thinkingmachines	1441.7 elo	77.0	100% confidence 100 percent, Full
49	Qwen3.8 27B Alibaba / Qwen	1440.7 elo	76.9	69% confidence 69 percent, Medium
50	Hy3 tencent	1440.7 elo	76.9	88% confidence 88 percent, High
51	GLM-4.6 Z.ai	1439.8 elo	76.8	55% confidence 55 percent, Medium
52	Qwen3.5 397B-A17B Alibaba / Qwen	1438.0 elo	76.7	69% confidence 69 percent, Medium
53	GLM-5V-Turbo Z.ai	1437.2 elo	76.6	32% confidence 32 percent, Low
54	Qwen3.6 Plus Alibaba / Qwen	1436.7 elo	76.6	80% confidence 80 percent, High
55	MiMo-V2-Pro xiaomi	1436.1 elo	76.5	75% confidence 75 percent, Medium
56	GLM-4.7 Z.ai	1435.5 elo	76.5	69% confidence 69 percent, Medium
57	Gemini 3.5 Flash Lite Google	1434.7 elo	76.4	100% confidence 100 percent, Full
58	Claude Sonnet 4.5 Anthropic	1434.7 elo	76.4	80% confidence 80 percent, High
59	Gemma 4 26B A4B IT Google	1433.9 elo	76.3	64% confidence 64 percent, Medium
60	DeepSeek V4 Flash DeepSeek	1432.1 elo	76.1	53% confidence 53 percent, Medium
61	MiniMax-M3 minimax	1432.1 elo	76.1	100% confidence 100 percent, Full
62	GPT-5.6 Luna OpenAI	1431.0 elo	76.0	100% confidence 100 percent, Full
63	GLM-4.5 Z.ai	1430.3 elo	76.0	47% confidence 47 percent, Low
64	Mistral Large 3 mistral	1427.6 elo	75.7	37% confidence 37 percent, Low
65	MiMo-V2.5 xiaomi	1427.5 elo	75.7	88% confidence 88 percent, High
66	Grok 4.6 xAI	1427.3 elo	75.7	100% confidence 100 percent, Full
67	MiMo-V2-Omni xiaomi	1423.4 elo	75.3	75% confidence 75 percent, Medium
68	GPT-5.1 OpenAI	1422.2 elo	75.2	85% confidence 85 percent, High
69	Mistral Medium 3.5 mistral	1421.2 elo	75.1	37% confidence 37 percent, Low
70	DeepSeek V3.2 DeepSeek	1420.1 elo	75.0	56% confidence 56 percent, Medium
71	Qwen3 VL 235B A22B Instruct Alibaba / Qwen	1420.1 elo	75.0	32% confidence 32 percent, Low
72	DeepSeek-V3.1 DeepSeek	1419.7 elo	75.0	32% confidence 32 percent, Low
73	GPT-5.5 Instant OpenAI	1419.5 elo	75.0	64% confidence 64 percent, Medium
74	Qwen3 235B-A22B Instruct 2507 Alibaba / Qwen	1419.3 elo	75.0	48% confidence 48 percent, Low
75	Claude Opus 4.1 Anthropic	1418.6 elo	74.9	80% confidence 80 percent, High
76	Qwen3-Next 80B-A3B Instruct Alibaba / Qwen	1417.3 elo	74.8	32% confidence 32 percent, Low
77	Qwen3.5 122B-A10B Alibaba / Qwen	1417.1 elo	74.8	37% confidence 37 percent, Low
78	Gemini 2.5 Flash Google	1417.0 elo	74.7	47% confidence 47 percent, Low
79	Gemini 3.1 Flash Lite Preview Google	1415.7 elo	74.6	48% confidence 48 percent, Low
80	Kimi K2 Thinking Turbo Moonshot AI	1414.4 elo	74.5	64% confidence 64 percent, Medium
81	Inkling Small thinkingmachines	1413.7 elo	74.4	100% confidence 100 percent, Full
82	Qwen3 Max Alibaba / Qwen	1412.4 elo	74.3	69% confidence 69 percent, Medium
83	GPT-5.2 OpenAI	1412.3 elo	74.3	100% confidence 100 percent, Full
84	GPT-5.4 mini OpenAI	1411.4 elo	74.2	100% confidence 100 percent, Full
85	o3 OpenAI	1409.9 elo	74.1	87% confidence 87 percent, High
86	Qwen3.5 27B Alibaba / Qwen	1408.9 elo	74.0	37% confidence 37 percent, Low
87	Grok 4.1 Fast (Reasoning) xAI	1408.0 elo	73.9	32% confidence 32 percent, Low
88	GPT-5 OpenAI	1406.1 elo	73.7	100% confidence 100 percent, Full
89	MiniMax-M2.7 minimax	1404.6 elo	73.6	88% confidence 88 percent, High
90	Step 3.5 Flash stepfun	1403.0 elo	73.4	88% confidence 88 percent, High
91	Qwen3 VL 235B A22B Thinking Alibaba / Qwen	1400.7 elo	73.2	32% confidence 32 percent, Low
92	Grok 4.7 xAI	1399.8 elo	73.1	100% confidence 100 percent, Full
93	Qwen3.5 Flash Alibaba / Qwen	1397.0 elo	72.8	64% confidence 64 percent, Medium
94	Grok 4.3 xAI	1396.7 elo	72.8	80% confidence 80 percent, High
95	Claude Haiku 4.5 Anthropic	1396.1 elo	72.7	64% confidence 64 percent, Medium
96	MiMo-V2-Flash xiaomi	1395.4 elo	72.7	75% confidence 75 percent, Medium
97	GPT-6 Sol OpenAI	1395.4 elo	72.7	100% confidence 100 percent, Full
98	Qwen3.5 35B-A3B Alibaba / Qwen	1394.4 elo	72.6	64% confidence 64 percent, Medium
99	GPT-6 Luna OpenAI	1391.5 elo	72.3	100% confidence 100 percent, Full
100	GPT-5.3 Chat (latest) OpenAI	1389.1 elo	72.0	32% confidence 32 percent, Low
101	Solar Pro 4 upstage	1386.2 elo	71.7	100% confidence 100 percent, Full
102	GLM-4.5-Air Z.ai	1383.8 elo	71.5	37% confidence 37 percent, Low
103	GPT-4.1 OpenAI	1383.0 elo	71.4	100% confidence 100 percent, Full
104	Nemotron 3 Super 120B A12B nvidia	1377.8 elo	70.9	100% confidence 100 percent, Full
105	GLM-4.6V Z.ai	1376.8 elo	70.8	32% confidence 32 percent, Low
106	DeepSeek V3 0324 DeepSeek	1374.9 elo	70.6	74% confidence 74 percent, Medium
107	GPT-5 Mini OpenAI	1373.2 elo	70.4	99% confidence 99 percent, High
108	DeepSeek-R1 DeepSeek	1372.7 elo	70.3	88% confidence 88 percent, High
109	GPT-5.4 nano OpenAI	1372.1 elo	70.3	100% confidence 100 percent, Full
110	Qwen3-Next 80B-A3B (Thinking) Alibaba / Qwen	1368.3 elo	69.9	32% confidence 32 percent, Low
111	Qwen Max Alibaba / Qwen	1366.6 elo	69.7	75% confidence 75 percent, Medium
112	Qwen3 235B-A22B Alibaba / Qwen	1366.6 elo	69.7	76% confidence 76 percent, Medium
113	Claude Opus 4 Anthropic	1365.7 elo	69.6	100% confidence 100 percent, Full
114	GPT OSS 120B OpenAI	1365.4 elo	69.6	79% confidence 79 percent, Medium
115	Nova 2 Lite amazon	1361.8 elo	69.2	100% confidence 100 percent, Full
116	MiniMax-M2.5 minimax	1359.3 elo	68.9	100% confidence 100 percent, Full
117	Gemma 3 27B IT Google	1357.8 elo	68.8	74% confidence 74 percent, Medium
118	Qwen3-Coder 480B-A35B Instruct Alibaba / Qwen	1356.3 elo	68.6	47% confidence 47 percent, Low
119	o4-mini OpenAI	1353.2 elo	68.3	87% confidence 87 percent, High
120	GLM-4.7-Flash Z.ai	1350.9 elo	68.0	69% confidence 69 percent, Medium
121	GPT-4.1 mini OpenAI	1340.4 elo	66.9	87% confidence 87 percent, High
122	Qwen3 32B Alibaba / Qwen	1340.1 elo	66.8	67% confidence 67 percent, Medium
123	Trinity Large Thinking arcee-ai	1340.0 elo	66.8	100% confidence 100 percent, Full
124	MiniMax-M2 minimax	1339.9 elo	66.8	96% confidence 96 percent, High
125	Claude Sonnet 4 Anthropic	1339.1 elo	66.7	100% confidence 100 percent, Full
126	Trinity Large Preview arcee-ai	1338.7 elo	66.7	100% confidence 100 percent, Full
127	Mistral Small 3.2 mistral	1338.3 elo	66.6	32% confidence 32 percent, Low
128	Llama 3.3 Nemotron Super 49B v1.5 nvidia	1338.0 elo	66.6	100% confidence 100 percent, Full
129	Gemma 3 12B IT Google	1334.2 elo	66.2	64% confidence 64 percent, Medium
130	GLM-4.5V Z.ai	1332.8 elo	66.0	37% confidence 37 percent, Low
131	DeepSeek-V3 DeepSeek	1332.5 elo	66.0	93% confidence 93 percent, High
132	Command A cohere	1330.9 elo	65.8	75% confidence 75 percent, Medium
133	QwQ 32B Alibaba / Qwen	1329.2 elo	65.6	67% confidence 67 percent, Medium
134	GPT-5 Nano OpenAI	1320.0 elo	64.6	85% confidence 85 percent, High
135	o3-mini OpenAI	1319.4 elo	64.5	96% confidence 96 percent, High
136	Qwen3 30B A3B Alibaba / Qwen	1316.6 elo	64.2	64% confidence 64 percent, Medium
137	GPT-4o (2024-05-13) OpenAI	1300.4 elo	62.3	93% confidence 93 percent, High
138	Claude Sonnet 3.7 Anthropic	1299.5 elo	62.2	100% confidence 100 percent, Full
139	Claude Sonnet 3.5 v2 Anthropic	1297.7 elo	62.0	100% confidence 100 percent, Full
140	Gemma 3 4B IT Google	1290.8 elo	61.2	64% confidence 64 percent, Medium
141	GPT OSS 20B OpenAI	1287.6 elo	60.8	64% confidence 64 percent, Medium
142	GPT-4o mini OpenAI	1286.4 elo	60.6	100% confidence 100 percent, Full
143	GPT-4.1 nano OpenAI	1284.8 elo	60.4	82% confidence 82 percent, High
144	Llama 3.1 Nemotron 70B Instruct nvidia	1282.6 elo	60.2	100% confidence 100 percent, Full
145	GPT-4o (2024-08-06) OpenAI	1282.5 elo	60.2	100% confidence 100 percent, Full
146	Llama 4 Scout 17B Instruct Meta	1279.4 elo	59.8	64% confidence 64 percent, Medium
147	Mistral Small 3.1 24B mistral	1277.5 elo	59.6	32% confidence 32 percent, Low
148	Llama-3.3-70B-Instruct Meta	1274.0 elo	59.2	100% confidence 100 percent, Full
149	Mistral Large 2.1 mistral	1265.3 elo	58.1	100% confidence 100 percent, Full
150	Llama-3.1-70B-Instruct Meta	1260.9 elo	57.6	93% confidence 93 percent, High
151	Nova Pro amazon	1258.7 elo	57.3	100% confidence 100 percent, Full
152	Claude Haiku 3.5 Anthropic	1255.3 elo	56.9	100% confidence 100 percent, Full
153	Qwen2.5-Coder-32B-Instruct Alibaba / Qwen	1230.0 elo	53.7	90% confidence 90 percent, High
154	Command R+ cohere	1228.9 elo	53.6	75% confidence 75 percent, Medium
155	Nova Lite amazon	1228.5 elo	53.6	100% confidence 100 percent, Full
156	Aya Expanse 32B cohere	1224.0 elo	53.0	75% confidence 75 percent, Medium
157	Nova Micro amazon	1208.4 elo	51.1	100% confidence 100 percent, Full
158	Claude Haiku 3 Anthropic	1194.6 elo	49.3	93% confidence 93 percent, High
159	Command R cohere	1187.3 elo	48.4	75% confidence 75 percent, Medium
160	Llama-3.1-8B-Instruct Meta	1186.5 elo	48.3	93% confidence 93 percent, High
161	Aya Expanse 8B cohere	1185.1 elo	48.1	75% confidence 75 percent, Medium
162	Mistral Medium 3 mistral	1165.1 elo	45.6	85% confidence 85 percent, High
163	Llama-3.2-3B Meta	1109.5 elo	38.9	75% confidence 75 percent, Medium
164	Llama-3.2-1B Meta	1054.6 elo	32.6	93% confidence 93 percent, High

Results are as published by the source behind each value (hover or tap the number). Benchmark scores are shown individually for every source/variant (a model may have multiple rows), and vary by version, harness and date; the normalized column uses the method’s fixed 0–100 scales and feeds the preference pillar of the SI Score.