CMDB-1500 (Image): leaderboard

Surd AI CMDB-1500 fixed-input decision benchmark: accuracy on 300 image questions spanning candidate choice, yes/no, ordinal and multi-select decisions. The vendor-run leaderboard includes general-purpose LLMs and Surd AI decision models; text and image panels use separate fixed denominators.

Metric: Accuracy (%), fixed denominator of 300 questions (invalid answers count as wrong), effort as labelled. Source: surdai.com. Saturation forecast: Around February 2027. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Claude Sonnet 5.5Low92.67#24
2GPT-6.1 SolLow92#15
3GLM-5.3 FlashLow90.67#133
4GPT-6 LunaLow89.33#105
5SPX-CD-Pro88.67
6SPX-CD-Flash87.33
7Qwen 3.8 Flash83#125
8SPX-CD-Omni82.67
9Kimi K2.681.33#192
10DeepSeek V4.1 Flash79.67#67

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=cmdb-1500-image · How It Works · Data refreshed daily, snapshot 2026-11-21.