MODEL INTELLIGENCE

ModelRank

Compare models within one task and one source. Human preference, benchmark ability, and usage measure different things; their scores are not combined.

Source LMArena Text Style Control Source publication: Oct 28, 2026 · Days since publication: 6 Last successful source check: 2026-11-03 12:05 UTC Snapshot stored: 2026-11-03 04:05 UTC

Original source

Methodology and data notes

Primary mixed open/closed LLM ranking. Style control reduces preference bias caused by response formatting and tone.

Metric: Style-controlled Arena rating, higher is better

Ranking source

LMArena Text Style Control Style-controlled Arena rating LMArena Text Arena rating LMArena Search Style Control Style-controlled Arena rating

Rank	Model	Organization	Style-controlled Arena rating ↑	95% CI	Access	Evidence
1	Gemini 4 Carbon (High) Style-controlled Arena rating ↑	google	1,527.5	1518.1–1536.9	Proprietary	4,403 votes Source
2	Gemini 4 Argon (High) Style-controlled Arena rating ↑	google	1,520.3	1514.0–1526.5	Proprietary	11,889 votes Source
3	Claude Opus 5.5 (High) Style-controlled Arena rating ↑	anthropic	1,508.7	1502.1–1515.3	Proprietary	11,415 votes Source
4	Muse Wonder (Max) Style-controlled Arena rating ↑	meta	1,507.1	1499.8–1514.4	Proprietary	7,749 votes Source
5	Claude Fable 5.1 (Max) Style-controlled Arena rating ↑	anthropic	1,504.4	1498.6–1510.2	Proprietary	17,178 votes Source
6	Claude Fable 5 (High) Style-controlled Arena rating ↑	anthropic	1,503.9	1499.7–1508.1	Proprietary	45,341 votes Source
7	Claude Opus 4.6 (High) Style-controlled Arena rating ↑	anthropic	1,503.2	1499.9–1506.4	Proprietary	84,512 votes Source
8	Claude Opus 4.7 (High) Style-controlled Arena rating ↑	anthropic	1,501.0	1497.3–1504.8	Proprietary	65,110 votes Source
9	Claude Opus 4.6 Style-controlled Arena rating ↑	anthropic	1,496.1	1492.9–1499.3	Proprietary	89,324 votes Source
10	Gemini 3.8 Flash (High) Style-controlled Arena rating ↑	google	1,495.9	1491.5–1500.4	Proprietary	47,803 votes Source
11	Muse Spark 1.3 (Max) Style-controlled Arena rating ↑	meta	1,495.0	1489.8–1500.3	Proprietary	22,134 votes Source
12	Claude Opus 4.7 Style-controlled Arena rating ↑	anthropic	1,493.7	1490.0–1497.5	Proprietary	66,140 votes Source
13	Muse Spark 1.2 (xHigh) Style-controlled Arena rating ↑	meta	1,492.2	1485.8–1498.5	Proprietary	11,724 votes Source
14	Claude Opus 5 (High) Style-controlled Arena rating ↑	anthropic	1,491.6	1488.0–1495.1	Proprietary	73,167 votes Source
15	Claude Opus 5 (Max) Style-controlled Arena rating ↑	anthropic	1,490.5	1485.8–1495.2	Proprietary	35,531 votes Source
16	Muse Spark 1.1 Style-controlled Arena rating ↑	meta	1,490.2	1486.1–1494.4	Proprietary	46,354 votes Source
17	Gemini 3.7 Flash (High) Style-controlled Arena rating ↑	google	1,489.6	1484.9–1494.2	Proprietary	32,289 votes Source
18	Muse Spark Style-controlled Arena rating ↑	meta	1,489.0	1483.3–1494.8	Proprietary	14,140 votes Source
19	Kimi K3 (Max) Style-controlled Arena rating ↑	moonshot	1,488.2	1483.7–1492.6	Kimi K3 license	36,780 votes Source
20	Gemini 3.1 Pro Preview Style-controlled Arena rating ↑	google	1,486.9	1484.0–1489.8	Proprietary	132,547 votes Source
21	Gemini 3 Pro Style-controlled Arena rating ↑	google	1,485.4	1481.6–1489.2	Proprietary	42,057 votes Source
22	GPT-5.6 Sol (xHigh) Style-controlled Arena rating ↑	openai	1,484.9	1480.7–1489.1	Proprietary	43,796 votes Source
23	Gemini 3.6 Flash (High) Style-controlled Arena rating ↑	google	1,483.2	1478.7–1487.6	Proprietary	36,028 votes Source
24	Qwen3.8-Max Style-controlled Arena rating ↑	alibaba	1,482.4	1477.6–1487.2	Proprietary	31,915 votes Source
25	Claude Opus 4.8 (High) Style-controlled Arena rating ↑	anthropic	1,481.7	1477.8–1485.6	Proprietary	63,994 votes Source
26	GPT-5.5 (High) Style-controlled Arena rating ↑	openai	1,481.7	1478.0–1485.4	Proprietary	69,377 votes Source
27	GPT-6.1 Sol (Max) Style-controlled Arena rating ↑	openai	1,481.3	1474.3–1488.4	Proprietary	9,367 votes Source
28	GLM-5.3 (Max) Style-controlled Arena rating ↑	zai	1,480.6	1475.6–1485.5	MIT	26,606 votes Source
29	MiMo-V2.6-Pro Style-controlled Arena rating ↑	xiaomi	1,479.7	1470.3–1489.1	MIT	4,037 votes Source
30	Gemini 3.5 Flash (High) Style-controlled Arena rating ↑	google	1,477.6	1473.7–1481.5	Proprietary	48,376 votes Source
31	Claude Sonnet 5.5 (xHigh) Style-controlled Arena rating ↑	anthropic	1,477.1	1470.4–1483.9	Proprietary	10,033 votes Source
32	GPT-5.5 Style-controlled Arena rating ↑	openai	1,476.7	1473.0–1480.3	Proprietary	70,949 votes Source
33	GLM-5.3-Flash Style-controlled Arena rating ↑	zai	1,476.5	1471.9–1481.2	MIT	41,642 votes Source
34	Gemini 3.5 Flash (Medium) Style-controlled Arena rating ↑	google	1,476.4	1472.3–1480.6	Proprietary	47,196 votes Source
35	DeepSeek-V4.1-Flash-20260910 (Max) Style-controlled Arena rating ↑	deepseek	1,476.2	1470.1–1482.3	MIT	13,657 votes Source
36	GPT-5.2 Chat (20260210) Style-controlled Arena rating ↑	openai	1,475.7	1471.6–1479.7	Proprietary	35,974 votes Source
37	GPT-6 Astra (Max) Style-controlled Arena rating ↑	openai	1,475.6	1469.7–1481.6	Proprietary	15,038 votes Source
38	GLM-5.2 (Max) Style-controlled Arena rating ↑	zai	1,475.3	1471.2–1479.4	MIT	53,702 votes Source
39	GPT-5.4 (High) Style-controlled Arena rating ↑	openai	1,475.3	1471.5–1479.0	Proprietary	64,748 votes Source
40	Grok 4.20 Beta 1 Style-controlled Arena rating ↑	xai	1,474.6	1470.0–1479.2	Proprietary	27,889 votes Source
41	Claude Opus 4.8 Style-controlled Arena rating ↑	anthropic	1,474.4	1470.6–1478.3	Proprietary	65,121 votes Source
42	Qwen3.7-Max-Preview Style-controlled Arena rating ↑	alibaba	1,473.7	1464.0–1483.4	Proprietary	3,932 votes Source
43	Claude Opus 4.5 (High, 32K, 20251101) Style-controlled Arena rating ↑	anthropic	1,473.7	1469.9–1477.5	Proprietary	37,484 votes Source
44	GPT-5.5 Instant Style-controlled Arena rating ↑	openai	1,473.2	1468.2–1478.2	Proprietary	27,128 votes Source
45	Gemini 3 Flash Style-controlled Arena rating ↑	google	1,472.7	1468.4–1477.1	Proprietary	31,301 votes Source
46	Claude Sonnet 4.6 Style-controlled Arena rating ↑	anthropic	1,472.0	1468.4–1475.5	Proprietary	70,916 votes Source
47	Grok 4.20 Beta (Reasoning, 0309) Style-controlled Arena rating ↑	xai	1,471.7	1468.1–1475.4	Proprietary	66,422 votes Source
48	Grok 4.20 Multi-Agent Beta (0309) Style-controlled Arena rating ↑	xai	1,470.7	1467.0–1474.4	Proprietary	64,917 votes Source
49	Claude Opus 4.5 (20251101) Style-controlled Arena rating ↑	anthropic	1,469.7	1466.5–1472.8	Proprietary	72,825 votes Source
50	GPT-5.6 Terra (xHigh) Style-controlled Arena rating ↑	openai	1,467.3	1463.2–1471.5	Proprietary	48,166 votes Source