Inference
LiveBench

LiveBench Reasoning

Mean score over web-of-lies, zebra puzzles, spatial reasoning and navigation logic, on questions written after the models were trained.

ModelsModels with a result.55
Top resultBest result on this test.92.7%GPT-6 Astra Max
Top-3 spreadPoints from first to third.1.0 pts
UpdatedDate the source changed the results.25 Jun 2026

Result and cost

020406080100$0.001$0.003$0.01$0.03$0.1$0.3$1RESULTUS DOLLARS PER TASK · LOG SCALE

Results

55 results
#ModelResultThe score from the source.CostUS dollars to run one task.
01GPT-6 Astra Max92.7%$0.39
02Claude Opus 5.5max effort92.2%$0.252
03Claude Fable 5.1max effort91.7%$0.46
04GPT-5.6 Sol Max91.7%$0.285
05Claude Opus 5max effort91.2%$0.168
06GPT-5.6 Terra Max90.6%$0.234
07Grok 4.690.5%$0.087
08Smaug Agentic90.3%$0.189
09Muse Spark 1.2xhigh effort90.0%$0.099
10Claude Fable 5max effort89.7%$0.602
10Muse Spark 1.3xhigh effort89.7%$0.136
10GPT-5.5xhigh effort89.7%$0.289
13Gemini 3.8 Flashhigh effort89.3%$0.148
14Claude Opus 4.8max effort89.2%$0.606
15Claude Sonnet 5xhigh effort88.7%$0.383
16Claude Opus 4.6 Thinking Autohigh effort88.7%$0.277
17Qwen3.8 Max88.2%$0.131
18GPT-5.4xhigh effort88.1%$0.213
19Gemini 3.7 Flashhigh effort87.8%$0.069
20Muse Spark 1.1xhigh effort87.7%$0.075
21Qwen3.8-Flash-Next87.4%$0.022
22Claude Opus 4.7xhigh effort87.2%$0.303
23Grok 4.587.2%$0.088
24DeepSeek V4.1 Flash Max86.7%$0.023
25Smaug Flash86.2%<$0.01
26GLM-5.385.8%$0.265
27GPT-5.6 Luna Max85.6%$0.138
28DeepSeek V4 Flash Vision Exp85.4%$0.042
29Gemini 3.6 Flashhigh effort85.1%$0.097
30Claude Sonnet 4.6 Thinking Automedium effort84.8%$0.207
31DeepSeek V4 Pro 042384.3%$0.041
32Qwen3.7 Max83.3%$0.102
33GPT-5.2high effort83.2%$0.114
34Kimi K2.7 Code82.8%$0.053
34Smaug Mini82.8%$0.019
36Grok 4.7xhigh effort82.7%$0.385
37Gemini 3.5 Flashhigh effort82.0%$0.149
38GPT-5.4 nanoxhigh effort81.1%$0.044
39Union Alpha80.8%$0
40Claude Opus 4.5 Thinking 64Khigh effort80.1%$0.499
41Qwen3.8-27B80.0%$0.016
42Kimi K2.6 Thinking79.4%$0.089
43GLM-5.278.6%$0.121
44DeepSeek V4 Flash 042378.6%$0.014
45Inklingxhigh effort78.3%$0.121
46GPT-5.2-Codex77.7%$0.169
47GLM-5.3-Flash77.6%$0.014
48Ox Alpha Max76.6%$0
49Grok Build 0.176.4%<$0.01
50Qwen3.6 Plus75.8%$0.094
51MiniMax M374.5%$0.019
52GPT-5.4 minixhigh effort71.3%$0.157
53Grok 4.370.8%$0.041
54Qwen3.6-27B70.3%$0.133
55Gemini 3.5 Flash-Litehigh effort60.2%$0.031