benchgap
General

BabaIsBench leaderboard

As of 2026-11-19, the highest measured score on BabaIsBench is 100.0% by Claude Fable 5. 21 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5100.0%measured
2Claude Fable 5.1100.0%measured
3Claude Fable 5.2100.0%estimated ± 1.2 pp, medium confidence
4Claude Opus 5100.0%measured
5Claude Opus 5.5100.0%estimated ± 1.2 pp, medium confidence
6Gemini 3.7 Flash100.0%measured
7Gemini 3.8 Flash100.0%measured
8Gemini 4 Argon100.0%estimated ± 1.2 pp, high confidence
9Gemini 4 Carbon100.0%estimated ± 1.2 pp, medium confidence
10Gemini 4 Hydrogen100.0%estimated ± 1.2 pp, high confidence
11GPT-5.5100.0%measured
12GPT-5.6 Sol100.0%measured
13GPT-6.1 Sol100.0%estimated ± 1.2 pp, high confidence
14GPT-6.1 Terra100.0%estimated ± 1.2 pp, medium confidence
15GPT-6.2 Astra100.0%estimated ± 1.2 pp, medium confidence
16GPT-6 Astra100.0%measured
17GPT-6 Sol100.0%estimated ± 1.2 pp, high confidence
18Claude Haiku 5.599.3%estimated ± 1.2 pp, high confidence
19Muse Wonder97.8%estimated ± 1.2 pp, high confidence
20GPT-5.496.6%estimated ± 1.2 pp, high confidence
21Claude Sonnet 5.596.5%estimated ± 14.1 pp, low confidence
22Claude Opus 4.896.0%measured
23GLM-5.296.0%measured
24GLM-5.396.0%measured
25Grok 4.696.0%measured
26Kimi K396.0%measured
27GPT-5.295.9%estimated ± 1.2 pp, high confidence
28Claude Opus 4.794.5%estimated ± 1.2 pp, high confidence
29Gemini 3.1 Pro92.0%measured
30Claude Opus 4.591.8%estimated ± 1.2 pp, high confidence
31Claude Opus 4.690.6%estimated ± 1.2 pp, medium confidence
32DeepSeek V4.1 Flash 101789.0%estimated ± 14.1 pp, low confidence
33Gemini 3.5 Flash88.0%measured
34Gemini 3.6 Flash88.0%measured
35GPT-5.6 Terra88.0%measured
36GPT-6 Luna86.2%estimated ± 14.1 pp, low confidence
37Muse Spark 1.276.3%estimated ± 14.1 pp, low confidence
38DeepSeek V4 Flash 073175.0%measured
39Grok 4.575.0%measured
40Claude Sonnet 567.0%measured
41GLM-5.3-Flash67.0%measured
42GPT-5.6 Luna63.0%measured
43Kimi K2.659.1%estimated ± 1.2 pp, medium confidence
44Qwen3.8 Max58.0%measured
45Hy354.0%measured
46MiniMax M346.0%measured
47Muse Spark 1.346.0%measured
48Qwen3.6-27B29.0%measured
49Qwen3.7 Max25.0%measured
50Qwen3.8-27B21.0%measured

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General