Atlas

Benchmarks

← All benchmarks

Arena Vision — Homework (No Style Control)

Multimodal

Raw Vision Arena rating for visually grounded homework prompts without style-control adjustment. Higher ratings indicate stronger preference in blind pairwise user votes.

Top models (higher is better)

ModelScore
Claude Fable 51361
GPT-6.1 Sol1356
Gemini 3.8 Flash1351
Claude Fable 5.11345
Claude Opus 51338
GPT-6 Astra1337
Opus 4.71337
Opus 4.81337
Opus 4.61337
GPT-5.41336
Qwen3.8-Max1336
GPT-5.51332
Claude Sonnet 5.51331
Gemini 3.5 Flash1330
Gemini 3.7 Flash1320
GLM-5.3 Flash1318
Gemini 3.1 Pro Preview1316
Gemini 3 Pro Preview1315
Gemini 3.6 Flash1314
Gemini 3 Flash Preview1313
GPT-5.6 Terra1312
Sonnet 4.61311
Sonnet 51311
Kimi K2.61310
Gemma 4 31B IT1310
Grok 4.71309
DeepSeek V4.1 Flash1307
GPT-5.6 Sol1305
Muse Spark1299
Kimi K2.51297
Muse Spark 1.31296
Grok 4.51295
Qwen3.7-Plus (source unspecified)1295
Qwen3.8 27B1295
Gemma 4 26B A4B IT1293
Qwen3.5 397B A17B1292
Gemini 3.5 Flash-Lite1291
Muse Spark 1.11287
GPT-5.4 Mini1285
Grok 4.61283
Loading Atlas data…