All KradleBench models
KradleBench17 models, sorted by Overall, highest first.
| Model | |||||||||
|---|---|---|---|---|---|---|---|---|---|
GPT-6 AstraOpenAI | 56.89 | 58.00 | 41.00 | 46.67 | 39.33 | 99.44 | $1.67 | 3.3 s | 365 |
GPT-6.1 SolOpenAI | 53.32 | 52.00 | 44.33 | 47.67 | 32.33 | 90.28 | $0.29 | 2.8 s | 365 |
GPT-6.1 TerraOpenAI | 50.13 | 58.00 | 41.33 | 41.33 | 34.67 | 75.33 | $0.18 | 2.0 s | 365 |
Claude Opus 5.5Anthropic | 44.66 | 61.00 | 26.00 | 27.00 | 30.00 | 79.30 | $1.07 | 3.8 s | 365 |
Gemini 4 ArgonGoogle | 36.66 | 31.00 | 26.67 | 19.33 | 28.67 | 77.62 | $0.58 | 6.2 s | 365 |
Claude Sonnet 5.5Anthropic | 34.99 | 46.00 | 24.00 | 26.66 | 30.00 | 48.28 | $0.59 | 2.3 s | 365 |
Claude Fable 5.1Anthropic | 34.78 | 39.66 | 21.33 | 18.33 | 30.00 | 64.60 | $2.35 | 4.1 s | 365 |
Muse Spark 1.3Meta | 21.69 | 36.67 | 12.00 | 14.33 | 2.67 | 42.76 | $0.50 | 3.4 s | 365 |
Gemini 3.8 FlashGoogle | 21.14 | 24.67 | 18.00 | 7.00 | 2.00 | 54.03 | $0.21 | 5.7 s | 365 |
DeepSeek V4.1 FlashDeepSeek | 20.31 | 39.00 | 13.66 | 24.00 | 3.33 | 21.58 | $0.069 | 2.4 s | 365 |
GPT-6 LunaOpenAI | 16.53 | 13.67 | 13.00 | 18.33 | 0.67 | 36.96 | $0.024 | 4.1 s | 365 |
Muse Spark OpenMeta | 16.03 | 9.67 | 19.67 | 12.00 | 1.33 | 37.49 | $0.52 | 4.4 s | 365 |
MiniMax M3.1 FlashMiniMax | 12.58 | 18.00 | 24.33 | 13.00 | 0.00 | 7.55 | $0.052 | 2.5 s | 365 |
Grok 4.7xAI | 12.00 | 34.33 | 16.34 | 9.33 | 0.00 | 0.00 | $0.66 | 3.2 s | 365 |
Gemma 4 31BGoogle | 10.73 | 17.66 | 16.67 | 12.00 | 0.00 | 7.30 | $0.056 | 4.6 s | 365 |
Nemotron 3.5 Super VLNVIDIA | 9.36 | 13.33 | 13.67 | 11.67 | 0.00 | 8.14 | $0.062 | 2.7 s | 365 |
GLM 5.3Z.ai | 6.71 | 16.00 | 14.34 | 1.33 | 0.00 | 1.86 | $0.12 | 9.6 s | 365 |
Overall weights the five dimensions equally. Cost is USD per run; latency is median model response time.