Inference
EN
LiveBench

LiveBench Language

Mean score over Connections word puzzles, unscrambling film plot summaries and repairing typos. English only. Not in LiveBench 2.0.

ModelsModels with a result.66
Top resultBest result on this benchmark.90.7%Claude Fable 5
Top-3 spreadPoints from first to third.1.2 pts
UpdatedDate the source changed the results.25 Jun 2026

Result and cost

65 of 66 results

Results

66 results
#ModelResultThe score from the source.CostUS dollars to run one task.
01Claude Fable 5max effort90.7%$0.446
02GPT-6.1 Solmax effort90.1%$0.041
03Claude Fable 5.1max effort89.5%$0.218
04GPT-6 Astramax effort89.4%$0.222
05Claude Opus 5max effort88.7%$0.096
06Gemini 3.8 Flashhigh effort87.8%$0.088
07GPT-5.6 Solmax effort87.7%$0.179
08GPT-5.5xhigh effort87.4%$0.19
09Claude Opus 5.5max effort86.3%$0.284
10Kimi K385.5%$0.147
10Gemini 3.7 Flashhigh effort85.5%$0.031
12Gemini 3.1 Pro Previewhigh effort85.4%$0.097
13GPT-6 Solmax effort85.3%$0.086
14Gemini 3.5 Flashhigh effort84.6%$0.059
15Smaug Agentic84.4%$0.139
16Gemini 3.6 Flashhigh effort83.9%$0.051
17Grok 4.683.7%$0.035
18Claude Opus 4.6 Thinking Autohigh effort83.3%$0.101
19Muse Wondermax effort82.9%$0.238
19GPT-5.6 Terramax effort82.9%$0.173
21Grok 4.582.8%$0.036
21Muse Spark 1.3xhigh effort82.8%$0.112
23GPT-5.4xhigh effort82.6%$0.19
24DeepSeek V4 Pro 081382.1%$0.021
25Claude Opus 4.5 Thinking 64Khigh effort81.3%$0.186
26DeepSeek V4.1 Flash 1017max effort81.2%$0.017
27DeepSeek V4 Flash Vision Exp80.4%$0.03
28Grok 4.7xhigh effort80.1%$0.305
29GLM-5.379.9%$0.241
30GPT-5.2high effort79.8%$0.042
31Qwen3.7 Max79.7%$0.037
31Qwen3.8 Max79.7%$0.135
31Claude Opus 4.8max effort79.7%$0.486
34Smaug Flash79.2%$0.0062
34DeepSeek V4 Flash 073179.2%$0.0036
36Muse Spark 1.2xhigh effort78.6%$0.056
37DeepSeek V4 Pro 042378.1%$0.014
38Claude Sonnet 5.5max effort78.0%$0.316
39Claude Opus 4.7xhigh effort77.9%$0.097
39Kimi K2.7 Code77.9%$0.035
41GLM-5.3-Flash77.3%$0.0065
42Smaug Mini77.0%$0.018
43MiniMax M376.8%$0.011
44GLM-5.276.2%$0.052
45Claude Sonnet 4.6 Thinking Automedium effort76.1%$0.061
46Kimi K2.6 Thinking75.1%$0.068
47Qwen3.6 Plus75.0%$0.032
47Claude Sonnet 5xhigh effort75.0%$0.127
49Qwen3.8-Flash-Next74.6%$0.025
50Qwen3.8-27B74.3%$0.032
50Muse Spark 1.1xhigh effort74.3%$0.045
52GPT-6 Lunamax effort73.8%$0.014
53GPT-5.2-Codex73.7%$0.071
54Grok 4.373.6%$0.023
55Inklingxhigh effort73.5%$0.085
56GPT-5.6 Lunamax effort72.6%$0.097
57Grok Build 0.172.5%$0.0012
58Gemini 3.5 Flash-Litehigh effort71.8%$0.016
59GPT-5.4 minixhigh effort71.0%$0.13
60DeepSeek V4 Flash 042370.1%$0.005
61Ox Alpha Max66.1%$0
62Qwen3.6-27B63.3%$0.048
63Claude Haiku 5.5max effort62.6%$0.022
64GPT-5.4 nanoxhigh effort62.5%$0.037
65Mistral Large 4high effort49.6%$0.053
66Nemotron 3.5 Lightning 30B A3B38.3%$0.014