Inference
EN
LiveBench

LiveBench Research

Research a brief and write a report with sources. A rubric for each task grades the report.

ModelsModels with a result.52
Top resultBest result on this benchmark.88.9%Claude Opus 5
Top-3 spreadPoints from first to third.2.0 pts
UpdatedDate the source changed the results.14 Oct 2026

Result and cost

52 of 52 results

Results

52 results
#ModelResultThe score from the source.CostUS dollars to run one task.
01Claude Opus 5max effort88.9%$16.97
02Muse Wondermax effort88.0%$3.51
03Claude Fable 5max effort86.9%$17.27
04DeepSeek V4.1 Flash 1017max effort86.8%$0.409
05Grok 4.7xhigh effort86.4%$4.58
06Grok 4.5xhigh effort85.9%$0.818
07GPT-6 Astramax effort85.2%$17.66
08Claude Opus 5.5max effort85.0%$8.82
09Grok 4.6xhigh effort84.5%$2.33
10Claude Sonnet 5.5max effort83.6%$8.19
11GPT-6.1 Solmax effort83.5%$2.42
12Gemini 4 Argonhigh effort83.1%$5.96
13Claude Sonnet 5max effort82.8%$5.96
14Claude Fable 5.1max effort82.0%$20.89
15Claude Haiku 5.5max effort81.9%$1.94
16Claude Opus 4.8max effort81.8%$6.24
17Muse Spark Openmax effort80.3%$0.732
18Mistral Large 4.0high effort79.7%$0.823
19DeepSeek V4 Flash 0731max effort79.2%$0.211
20Gemini 3.8 Flashhigh effort79.1%$2.64
21Kimi K3max effort79.0%$2.40
22MiMo-V2.6-Prohigh effort76.7%$0.142
23Gemini 3.7 Flashhigh effort76.5%$2.68
24Qwen3.8-Flash-Nextmax effort74.6%$0.317
25GPT-6.1 Terramax effort73.9%$1.33
26GLM-5.3max effort73.2%$1.63
27Step 5 Previewmax effort72.8%$1.23
27Beam 1.0max effort72.8%$2.39
29MiniMax M3.1 Flashmax effort72.7%$0.266
30Nemotron Super VL-3.5max effort70.9%$0.167
30Nemotron Super VL-3.5 Nvfp4max effort70.9%$0.10
32GLM-5.3-Flashmax effort70.1%$0.187
33Qwen3.8 Maxmax effort69.9%$1.95
34GPT-6 Solmax effort69.3%$3.51
35DeepSeek V4 Flash Vision Expmax effort69.2%$0.249
36Muse Spark 1.3max effort68.6%$2.69
37MiniMax M3max effort68.0%$0.252
38Smaug Nemotron 3.5 Nvfp4max effort67.8%$0.093
39Muse Spark 1.2xhigh effort67.7%$1.78
40Gemini 3.1 Pro Previewhigh effort65.7%$0.77
41Claude Sonnet 4.6max effort63.6%$3.90
42GPT-5.5xhigh effort61.6%$17.14
43GPT-5.6 Terramax effort60.4%$7.72
44Inkling59.9%$1.67
45Qwen3.8-27Bmax effort59.4%$1.13
46GPT-5.6 Solmax effort54.5%$14.43
47GPT-5.6 Lunamax effort51.0%$0.821
48Gemma 4 31Bmax effort44.4%$0.038
49GPT-6 Lunamax effort43.5%$0.355
50Claude Haiku 4.5max effort42.2%$0.432
51Nemotron 3.5 Lightning 30B A3Bmax effort40.0%$0.087
52Mistral Large 4high effort28.1%$2.33