Atlas

Benchmarks

← All benchmarks

Vision Arena Overall

Multimodal · 2026-06-10

Arena-style leaderboard for multimodal AI models capable of reasoning over visual inputs, reported as arena/Elo-style model scores.

Top models (higher is better)

ModelScore
Claude Fable 51318
Opus 4.71303
Opus 4.61299
Claude Opus 51295
Muse Spark1294
Gemini 3.6 Flash1290
Gemini 3 Pro Preview1289
GPT-5.51287
Opus 4.81286
Gemini 3.5 Flash1284
Muse Spark 1.11283
Grok 4.51282
GPT-5.41281
GPT-5.6 Sol1280
Gemini 3.1 Pro Preview1278
Sonnet 51275
Sonnet 4.61275
Gemini 3 Flash Preview1271
GPT-5.6 Terra1269
Kimi K2.61264
Qwen3.7-Plus1262
Gemma 4 31B1257
Gemini 3.5 Flash-Lite1257
GPT-5.4 Mini1255
GPT-5.11250
Kimi K2.51250
Qwen3.5 397B A17B1248
GPT-5.6 Luna1248
Gemini 2.5 Pro1246
GPT-5.21244
Grok 4.31244
MiMo-V2.51240
MiniMax M31239
Gemma 4 26B A4B1239
Gemini 3.1 Flash-Lite Preview1234
GLM 5V Turbo1230
Qwen3.5 122B A10B1227
Gemini 2.5 Flash Preview (09-2025)1225
MiMo-V2-Omni1217
O3 (2025-04-16)1217
Loading Atlas data…