Significance bands

Adjacent ranks that the measurement cannot separate. The band is the object; the integer is a footnote. We do not invent a crisp #1.

94% adjacent pairs inside published error

72/145 distinct significance ranks / scored models

4 joint first among listed models

LMArena publishes per-model 95% CI half-widths (CC BY 4.0). Adjacent models are inseparable when their intervals overlap.

Sig. rank	Model	Elo	± half-width
1	Claude Opus 5.5 Check	1512	9.1
1	Claude Fable 5.1 Check	1511	6.4
1	Claude Opus 4.6 Check	1503	3.4
1	Claude Opus 5 Check	1502	3.8
3	Gemini 3.8 Flash Check	1497	5.0
5	Claude Fable 5 Check	1492	4.4
3	MiMo-V2.6-Pro Check	1491	9.4
5	Claude Opus 4.7 Check	1490	3.8
5	Muse Spark 1.3 Check	1490	6.2
5	Gemini 3.7 Flash Check	1487	5.2
5	Muse Spark 1.2 Check	1483	9.1
9	Gemini 3.1 Pro Preview Check	1480	3.0
8	Gemini 3.6 Flash Check	1480	4.4
9	Muse Spark 1.1 Check	1479	4.4
10	Gemini 3.5 Flash Check	1478	4.1
11	Kimi K3 Check	1476	4.8
12	GLM 5.3 Check	1471	5.5
14	GLM 5.2 Check	1470	4.3
13	GLM 5.3 Flash Check	1470	5.3
17	GPT-5.5 Check	1467	3.7
12	Claude Sonnet 5.5 Check	1467	10.4
17	MiMo-V2.5-Pro Check	1465	3.6
17	DeepSeek V4.1 Flash Check	1463	6.9
19	GLM 5.1 Check	1462	3.7
19	Claude Opus 4.8 Check	1461	3.9
22	Claude Sonnet 4.6 Check	1458	3.5
22	Gemini 2.5 Pro Check	1458	2.4
20	MiMo-V2.6-Flash Check	1456	7.8
22	GPT-5.6 Sol Check	1456	4.4
22	Kimi K2.6 Check	1456	4.4
22	Qwen3.7 Plus Check	1455	4.3
25	GPT-5.4 Check	1452	3.7
28	Claude Opus 4.5 Check	1451	3.1
28	Grok 4.5 Check	1448	4.3
27	Qwen3.6 Max Preview Check	1447	8.2
30	GPT-5.6 Terra Check	1446	4.4
30	GLM 5 Check	1446	4.2
25	GPT-6.1 Sol Check	1446	10.5
29	DeepSeek V4 Pro 0423 Check	1445	6.6
33	Kimi K2.5 Check	1445	3.3
30	Gemma 4 31B Check	1443	7.4
34	Claude Sonnet 5 Check	1443	4.2
31	GPT-6 Astra Check	1442	6.8
34	Inkling Check	1442	4.5
34	Qwen3.8 27B Check	1441	5.4
34	Hy3 Check	1441	6.5
34	GLM 4.6 Check	1440	3.9
38	Claude Sonnet 4.5 Check	1439	2.9
38	Qwen3.5 397B A17B Check	1438	3.1
35	GLM 5V Turbo Check	1437	6.4
38	Qwen3.6 Plus Check	1437	4.1
38	GLM 4.7 Check	1436	6.0
38	Gemini 3.5 Flash Lite Check	1435	4.5
38	Gemma 4 26B A4B Check	1434	7.4
42	DeepSeek V4 Flash 0423 Check	1432	4.0
42	MiniMax M3 Check	1432	3.9
46	GPT-5.6 Luna Check	1431	4.4
47	GLM 4.5 Check	1430	4.9
50	R1 0528 Check	1428	5.7
51	MiMo-V2.5 Check	1428	4.2
51	Grok 4.6 Check	1427	5.1
56	DeepSeek V3.2 Check	1425	3.5
54	DeepSeek V3.2 Exp Check	1422	6.3
58	GPT-5.1 Check	1422	3.6
56	Mistral Medium 3.5 Check	1421	6.4
57	Qwen3 VL 235B A22B Instruct Check	1420	6.5
58	DeepSeek V3.1 Check	1420	6.0
62	Qwen3 235B A22B Instruct 2507 Check	1419	2.5
62	Claude Opus 4.1 Check	1419	3.0
62	Qwen3 Next 80B A3B Instruct Check	1417	4.8
56	DeepSeek V3.1 Terminus Check	1417	9.7
62	Qwen3.5-122B-A10B Check	1417	4.3
63	Gemini 2.5 Flash Check	1417	2.4
63	Gemini 3.1 Flash Lite Preview Check	1416	3.6
62	Qwen3 235B A22B Thinking 2507 Check	1415	6.5
64	Inkling Small Check	1414	4.9
63	Qwen3 Max Check	1412	6.5
67	GPT-5.2 Check	1412	3.1
67	GPT-5.4 Mini Check	1411	3.7
71	o3 Check	1410	3.6
71	Qwen3.5-27B Check	1409	4.3
74	GPT-5 Check	1406	4.4
77	MiniMax M2.7 Check	1405	3.4
79	Step 3.5 Flash Check	1403	3.5
78	Qwen3 VL 235B A22B Thinking Check	1401	6.8
78	Grok 4.7 Check	1400	7.8
84	Qwen3.5-Flash Check	1397	3.6
84	Grok 4.3 Check	1397	3.6
85	Claude Haiku 4.5 Check	1396	2.4
82	GPT-6 Sol Check	1395	6.9
85	Qwen3.5-35B-A3B Check	1394	4.2
85	GPT-6 Luna Check	1392	6.5
90	Solar Pro 4 Check	1386	5.7
91	Qwen3 30B A3B Instruct 2507 Check	1384	4.9
92	GLM 4.5 Air Check	1384	4.2
92	GPT-4.1 Check	1383	3.7
92	GLM 4.6V Check	1377	11.1
97	DeepSeek V3 0324 Check	1375	3.9
97	GPT-5 Mini Check	1373	4.6
97	R1 Check	1373	4.8
97	GPT-5.4 Nano Check	1372	3.8
97	Qwen3 Next 80B A3B Thinking Check	1368	5.9
97	Qwen3 235B A22B Check	1367	4.7
98	o1 Check	1366	4.5
98	gpt-oss-120b Check	1365	4.4
101	Nova 2 Lite Check	1362	6.0
102	MiniMax M2.5 Check	1359	3.9
104	Gemma 3 27B Check	1358	3.7
102	Mercury 2 Check	1355	10.3
106	o4 Mini Check	1353	4.0
106	GLM 4.7 Flash Check	1351	5.7
110	MiniMax M1 Check	1343	4.3
112	GPT-4.1 Mini Check	1340	4.3
109	Qwen3 32B Check	1340	9.3
112	Trinity Large Thinking Check	1340	4.6
110	MiniMax M2 Check	1340	7.7
112	Claude Sonnet 4 Check	1339	4.4
112	o3 Mini High Check	1337	5.3
112	Gemma 3 12B Check	1334	9.4
112	GLM 4.5V Check	1333	8.4
113	DeepSeek V3 Check	1333	4.7
120	GPT-5 Nano Check	1320	6.9
122	o3 Mini Check	1319	3.5
119	Granite 4.2 8B Check	1318	11.1
122	Qwen3 30B A3B Check	1317	4.7
126	GPT-4o (2024-05-13) Check	1300	3.4
126	Gemma 3 4B Check	1291	9.2
127	gpt-oss-20b Check	1288	6.4
127	GPT-4o-mini (2024-07-18) Check	1286	3.5
127	GPT-4.1 Nano Check	1285	7.6
127	GPT-4o (2024-08-06) Check	1283	4.2
131	Llama 3.3 70B Instruct Check	1274	3.5
132	GPT-4 Turbo Check	1272	3.9
132	Qwen2.5 72B Instruct Check	1269	4.1
133	Mistral Large 2407 Check	1266	3.9
135	Llama 3.1 70B Instruct Check	1261	3.7
137	Mistral Small 3 Check	1234	5.9
137	Gemma 2 27B Check	1231	3.4
137	Qwen2.5 Coder 32B Instruct Check	1230	8.0
137	Command R+ (08-2024) Check	1229	6.5
141	Phi 4 Check	1217	4.6
142	Command R (08-2024) Check	1187	6.5
142	Llama 3.1 8B Instruct Check	1187	4.1
144	Llama 3.2 3B Instruct Check	1110	7.5
145	Llama 3.2 1B Instruct Check	1055	7.7

Input hashes & method version

Read citation text and BibTeX

136 of 144 adjacent Arena pairs fall inside published intervals (94%). Standard-delivery models with measured confidence intervals only. As of 2026-10-06. Lens: LMArena Elo (CC BY 4.0). https://undominated.ai/significance/
Method: https://undominated.ai/methodology/
Arena data: https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset (CC BY 4.0). Derived calculations by Undominated.ai.
Publication: sha256:ecef83f767cd4d2c21766f6a5b4dffa36c06351f23657542d473bc1a030c0932; calculation SHA-256: 5ae4d0bee9b70298dd437b05df87c32acb407040b50bf88240677f295f25955d; permitted inputs: https://undominated.ai/data/publications/manifests/ecef83f767cd4d2c21766f6a5b4dffa36c06351f23657542d473bc1a030c0932.json

@misc{undominated-significance-2026-10-06,
  author = {{Undominated.ai}},
  title  = {significance evidence — Undominated.ai},
  year   = {2026},
  url    = {https://undominated.ai/significance/},
  note   = {as of 2026-10-06; LMArena Elo (CC BY 4.0)},
  annote = {Publication: sha256:ecef83f767cd4d2c21766f6a5b4dffa36c06351f23657542d473bc1a030c0932; calculation SHA-256: 5ae4d0bee9b70298dd437b05df87c32acb407040b50bf88240677f295f25955d; permitted inputs: https://undominated.ai/data/publications/manifests/ecef83f767cd4d2c21766f6a5b4dffa36c06351f23657542d473bc1a030c0932.json}
}

Choose models

Compare costs

Build with AI

Understand the numbers

Check our work

Reuse the data

Significance bands