Benchmark · Arenas: people's votes

LMArena Vision Arena

Which of two anonymous models' answers people prefer when the prompt comes with an image, fitted as a Bradley-Terry rating. Rows give the raw rating and the style-controlled one.

Score
arena rating (Bradley-Terry), higher is better
Run by
LMArena
Status
Current, newest result 19 November 2026
Updated
daily
Data
the published file, CC-BY-4.0, read 20 November 2026
Method
arena.ai/blog/leaderboard-changelog/

Results: vision style control

29 models we list, each at its best setting, best first.

#ModelScoreRangeWho ran itSettingDate
1Claude Fable 5.2 A13051287 to 1324Independenteffort max, style control on2026-11-19source
2Claude Opus 5.5 A12931284 to 1302Independenteffort high, style control on2026-11-19source
3Gemini 3.8 Flash AA12931285 to 1300Independenteffort high, style control on2026-11-19source
4GPT-6.1 Sol AA12901280 to 1300Independenteffort max, style control on2026-11-19source
5Claude Fable 5.1 A12901282 to 1298Independenteffort max, style control on2026-11-19source
6Muse Spark 1.3 D12881281 to 1296Independenteffort max, style control on2026-11-19source
7GPT-6 Astra A12811272 to 1290Independenteffort max, style control on2026-11-19source
8Gemini 3.1 Pro (preview) AA12801275 to 1285Independentstyle control on2026-11-19source
9GLM-5.3-Flash B12741268 to 1279Independentstyle control on2026-11-19source
10Gemini 3.5 Flash-Lite AA12661258 to 1273Independentstyle control on2026-11-19source
11Gemma 4 31B C12651259 to 1270Independentstyle control on2026-11-19source
12Claude Sonnet 5.5 A12641255 to 1274Independenteffort xhigh, style control on2026-11-19source
13GPT-6.1 Terra AA12631250 to 1276Independenteffort max, style control on2026-11-19source
14DeepSeek V4.1 Flash C12621252 to 1272Independenteffort max, style control on2026-11-19source
15GPT-6 Sol A12581249 to 1266Independenteffort max, style control on2026-11-19source
16Grok 4.7 BB12501242 to 1258Independenteffort xhigh, style control on2026-11-19source
17Step 5 Preview E12461236 to 1257Independentstyle control on2026-11-19source
18Qwen3.5-397B-A17B B12451240 to 1251Independentstyle control on2026-11-19source
19Qwen3.8-27B B12441237 to 1252Independentstyle control on2026-11-19source
20Gemma 4 26B A4B IT BB12411235 to 1248Independentstyle control on2026-11-19source
21Grok 4.3 BB12411236 to 1247Independentstyle control on2026-11-19source
22GPT-6 Luna A12341226 to 1242Independenteffort max, style control on2026-11-19source
23Mistral Large 4 C12331222 to 1244Independentstyle control on2026-11-19source
24MiniMax M3 C12311225 to 1237Independentstyle control on2026-11-19source
25Qwen3.5-122B-A10B B12261219 to 1232Independentstyle control on2026-11-19source
26Nemotron 3.5 Super VL C12111202 to 1220Independentprecision NVFP4, style control on2026-11-19source
27Inkling-Small B12081200 to 1216Independentstyle control on2026-11-19source
28Mistral Large 3 B12001193 to 1208Independentstyle control on2026-11-19source
29Mistral Medium 3.5 B11991189 to 1208Independentstyle control on2026-11-19source
Every setting, 30 rows
ModelAs the leaderboard names itScoreWho ran itSettingDate
Claude Fable 5.2Claude Fable 5.2 (Max)1305Independenteffort max, style control on2026-11-19source
Claude Opus 5.5Claude Opus 5.5 (High)1293Independenteffort high, style control on2026-11-19source
Gemini 3.8 FlashGemini 3.8 Flash (High)1293Independenteffort high, style control on2026-11-19source
GPT-6.1 SolGPT-6.1 Sol (Max)1290Independenteffort max, style control on2026-11-19source
Claude Fable 5.1Claude Fable 5.1 (Max)1290Independenteffort max, style control on2026-11-19source
Muse Spark 1.3Muse Spark 1.3 (Max)1288Independenteffort max, style control on2026-11-19source
Muse Spark 1.3Muse Spark 1.3 (xHigh)1288Independenteffort xhigh, style control on2026-11-19source
GPT-6 AstraGPT-6 Astra (Max)1281Independenteffort max, style control on2026-11-19source
Gemini 3.1 Pro (preview)Gemini 3.1 Pro Preview1280Independentstyle control on2026-11-19source
GLM-5.3-FlashGLM-5.3-Flash1274Independentstyle control on2026-11-19source
Gemini 3.5 Flash-LiteGemini 3.5 Flash-Lite1266Independentstyle control on2026-11-19source
Gemma 4 31BGemma 4 31B1265Independentstyle control on2026-11-19source
Claude Sonnet 5.5Claude Sonnet 5.5 (xHigh)1264Independenteffort xhigh, style control on2026-11-19source
GPT-6.1 TerraGPT-6.1 Terra (Max)1263Independenteffort max, style control on2026-11-19source
DeepSeek V4.1 FlashDeepSeek-V4.1-Flash-20260910 (Max)1262Independenteffort max, style control on2026-11-19source
GPT-6 SolGPT-6 Sol (Max)1258Independenteffort max, style control on2026-11-19source
Grok 4.7Grok 4.7 (xHigh)1250Independenteffort xhigh, style control on2026-11-19source
Step 5 PreviewStep 5 Preview1246Independentstyle control on2026-11-19source
Qwen3.5-397B-A17BQwen3.5-397B-A17B1245Independentstyle control on2026-11-19source
Qwen3.8-27BQwen3.8-27B1244Independentstyle control on2026-11-19source
Gemma 4 26B A4B ITGemma 4 26B A4B1241Independentstyle control on2026-11-19source
Grok 4.3Grok 4.31241Independentstyle control on2026-11-19source
GPT-6 LunaGPT-6 Luna (Max)1234Independenteffort max, style control on2026-11-19source
Mistral Large 4Mistral Large 41233Independentstyle control on2026-11-19source
MiniMax M3MiniMax-M31231Independentstyle control on2026-11-19source
Qwen3.5-122B-A10BQwen3.5-122B-A10B1226Independentstyle control on2026-11-19source
Nemotron 3.5 Super VLNVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4)1211Independentprecision NVFP4, style control on2026-11-19source
Inkling-SmallInkling-Small1208Independentstyle control on2026-11-19source
Mistral Large 3Mistral Large 31200Independentstyle control on2026-11-19source
Mistral Medium 3.5Mistral Medium 3.51199Independentstyle control on2026-11-19source

Results: vision

29 models we list, each at its best setting, best first.

#ModelScoreRangeWho ran itSettingDate
1Claude Fable 5.1 A13221314 to 1330Independenteffort max, style control off2026-11-19source
2Claude Opus 5.5 A13221313 to 1331Independenteffort high, style control off2026-11-19source
3Gemini 3.8 Flash AA13161309 to 1324Independenteffort high, style control off2026-11-19source
4Claude Fable 5.2 A13131295 to 1332Independenteffort max, style control off2026-11-19source
5Muse Spark 1.3 D13061299 to 1314Independenteffort max, style control off2026-11-19source
6Gemini 3.1 Pro (preview) AA12961292 to 1301Independentstyle control off2026-11-19source
6GLM-5.3-Flash B12961291 to 1302Independentstyle control off2026-11-19source
8Claude Sonnet 5.5 A12861277 to 1295Independenteffort xhigh, style control off2026-11-19source
9GPT-6.1 Sol AA12841274 to 1293Independenteffort max, style control off2026-11-19source
10Gemma 4 31B C12791274 to 1285Independentstyle control off2026-11-19source
11DeepSeek V4.1 Flash C12781268 to 1288Independenteffort max, style control off2026-11-19source
12GPT-6 Astra A12761268 to 1285Independenteffort max, style control off2026-11-19source
13Qwen3.8-27B B12711264 to 1278Independentstyle control off2026-11-19source
14Gemini 3.5 Flash-Lite AA12681261 to 1276Independentstyle control off2026-11-19source
15Step 5 Preview E12641254 to 1275Independentstyle control off2026-11-19source
16Qwen3.5-397B-A17B B12621257 to 1267Independentstyle control off2026-11-19source
17Gemma 4 26B A4B IT BB12591253 to 1266Independentstyle control off2026-11-19source
18MiniMax M3 C12481242 to 1255Independentstyle control off2026-11-19source
19GPT-6.1 Terra AA12471234 to 1260Independenteffort max, style control off2026-11-19source
20Mistral Large 4 C12461235 to 1257Independentstyle control off2026-11-19source
21Qwen3.5-122B-A10B B12441237 to 1251Independentstyle control off2026-11-19source
22Inkling-Small B12361228 to 1244Independentstyle control off2026-11-19source
23GPT-6 Sol A12321223 to 1241Independenteffort max, style control off2026-11-19source
24Grok 4.7 BB12311223 to 1239Independenteffort xhigh, style control off2026-11-19source
25Grok 4.3 BB12281222 to 1234Independentstyle control off2026-11-19source
26Mistral Large 3 B12221215 to 1230Independentstyle control off2026-11-19source
26Mistral Medium 3.5 B12221213 to 1232Independentstyle control off2026-11-19source
28Nemotron 3.5 Super VL C12131205 to 1222Independentprecision NVFP4, style control off2026-11-19source
29GPT-6 Luna A12101201 to 1218Independenteffort max, style control off2026-11-19source
Every setting, 30 rows
ModelAs the leaderboard names itScoreWho ran itSettingDate
Claude Fable 5.1Claude Fable 5.1 (Max)1322Independenteffort max, style control off2026-11-19source
Claude Opus 5.5Claude Opus 5.5 (High)1322Independenteffort high, style control off2026-11-19source
Gemini 3.8 FlashGemini 3.8 Flash (High)1316Independenteffort high, style control off2026-11-19source
Claude Fable 5.2Claude Fable 5.2 (Max)1313Independenteffort max, style control off2026-11-19source
Muse Spark 1.3Muse Spark 1.3 (Max)1306Independenteffort max, style control off2026-11-19source
Muse Spark 1.3Muse Spark 1.3 (xHigh)1300Independenteffort xhigh, style control off2026-11-19source
Gemini 3.1 Pro (preview)Gemini 3.1 Pro Preview1296Independentstyle control off2026-11-19source
GLM-5.3-FlashGLM-5.3-Flash1296Independentstyle control off2026-11-19source
Claude Sonnet 5.5Claude Sonnet 5.5 (xHigh)1286Independenteffort xhigh, style control off2026-11-19source
GPT-6.1 SolGPT-6.1 Sol (Max)1284Independenteffort max, style control off2026-11-19source
Gemma 4 31BGemma 4 31B1279Independentstyle control off2026-11-19source
DeepSeek V4.1 FlashDeepSeek-V4.1-Flash-20260910 (Max)1278Independenteffort max, style control off2026-11-19source
GPT-6 AstraGPT-6 Astra (Max)1276Independenteffort max, style control off2026-11-19source
Qwen3.8-27BQwen3.8-27B1271Independentstyle control off2026-11-19source
Gemini 3.5 Flash-LiteGemini 3.5 Flash-Lite1268Independentstyle control off2026-11-19source
Step 5 PreviewStep 5 Preview1264Independentstyle control off2026-11-19source
Qwen3.5-397B-A17BQwen3.5-397B-A17B1262Independentstyle control off2026-11-19source
Gemma 4 26B A4B ITGemma 4 26B A4B1259Independentstyle control off2026-11-19source
MiniMax M3MiniMax-M31248Independentstyle control off2026-11-19source
GPT-6.1 TerraGPT-6.1 Terra (Max)1247Independenteffort max, style control off2026-11-19source
Mistral Large 4Mistral Large 41246Independentstyle control off2026-11-19source
Qwen3.5-122B-A10BQwen3.5-122B-A10B1244Independentstyle control off2026-11-19source
Inkling-SmallInkling-Small1236Independentstyle control off2026-11-19source
GPT-6 SolGPT-6 Sol (Max)1232Independenteffort max, style control off2026-11-19source
Grok 4.7Grok 4.7 (xHigh)1231Independenteffort xhigh, style control off2026-11-19source
Grok 4.3Grok 4.31228Independentstyle control off2026-11-19source
Mistral Large 3Mistral Large 31222Independentstyle control off2026-11-19source
Mistral Medium 3.5Mistral Medium 3.51222Independentstyle control off2026-11-19source
Nemotron 3.5 Super VLNVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4)1213Independentprecision NVFP4, style control off2026-11-19source
GPT-6 LunaGPT-6 Luna (Max)1210Independenteffort max, style control off2026-11-19source

A rank is among the models we list with a result, not a place on the whole leaderboard, and a model offered by several hosts is shown once, on its own listing. These results don't move our grades. Data: LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset), CC BY 4.0, CC-BY-4.0.

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.