Atlas

Benchmarks

← All benchmarks

Arena Text — Hard Prompts (No Style Control)

Chat & Writing

Raw Arena rating for hard Text Arena prompts without style-control adjustment. Higher ratings indicate stronger preference in blind pairwise user votes.

Top models (higher is better)

ModelScore
Gemini 4 Argon1538
Claude Opus 5.51536
Claude Fable 5.11528
Claude Opus 51527
Opus 4.61524
MiMo-V2.6-Pro1510
Gemini 3.8 Flash1509
Claude Fable 51507
Opus 4.71506
Muse Wonder1505
Muse Spark 1.31504
Gemini 3.7 Flash1497
Qwen3.8-Max1497
Kimi K31495
Claude Sonnet 5.51493
GLM-5.3 Flash1491
Muse Spark 1.21490
GLM-5.31490
GPT-5.51489
Gemini 3.5 Flash1488
Gemini 3.6 Flash1486
MiMo-V2.5-Pro1486
Muse Spark 1.11485
GPT-5.41485
Gemini 3.1 Pro Preview1485
Sonnet 4.61485
GPT-5.6 Sol1484
DeepSeek V4.1 Flash1483
Qwen3.7 Max Preview1482
Opus 4.81482
ERNIE 5.11481
GLM-5.21480
Gemini 3 Pro Preview1480
Opus 4.51476
Muse Spark1474
MiMo-V2.6-Flash1474
GLM-5.11473
Kimi K2.61470
GPT-5.6 Terra1469
GPT-6.1 Sol1465
Loading Atlas data…