Inference
EN
LiveBench

LiveBench Long Agentic

Long agentic work: extend business apps over many turns, run a codebase through simulated months, build and tune systems. Graded by scripts.

ModelsModels with a result.51
Top resultBest result on this benchmark.85.0%Claude Opus 5.5
Top-3 spreadPoints from first to third.6.2 pts
UpdatedDate the source changed the results.14 Oct 2026

Result and cost

51 of 51 results

Results

51 results
#ModelResultThe score from the source.CostUS dollars to run one task.
01Claude Opus 5.5max effort85.0%$38.01
02Claude Fable 5.1max effort80.8%$51.57
03Claude Sonnet 5.5max effort78.8%$35.01
04GPT-6.1 Solmax effort73.0%$6.49
05Gemini 4 Argonhigh effort69.9%$13.84
06GPT-6 Solmax effort66.9%$8.62
07Claude Opus 5max effort62.1%$48.75
08Claude Haiku 5.5max effort57.6%$9.76
09Claude Fable 5max effort53.2%$59.74
09Muse Wondermax effort53.2%$14.38
11Grok 4.7xhigh effort50.1%$17.42
12GPT-5.6 Solmax effort48.5%$15.96
13GPT-5.6 Terramax effort48.4%$11.76
14Claude Opus 4.8max effort48.3%$51.97
15GPT-6 Lunamax effort48.1%$0.968
16MiMo-V2.6-Prohigh effort47.5%$1.23
17GPT-5.5xhigh effort44.6%$13.85
18Kimi K3max effort42.8%$17.99
19Qwen3.8 Maxmax effort42.3%$14.93
20Grok 4.6xhigh effort40.7%$9.05
21Muse Spark 1.3max effort40.4%$8.40
22DeepSeek V4.1 Flash 1017max effort40.1%$1.65
23GPT-6.1 Terramax effort39.6%$4.75
24Claude Sonnet 5max effort39.0%$33.21
24DeepSeek V4 Flash Vision Expmax effort39.0%$1.04
26Qwen3.8-Flash-Nextmax effort37.1%$1.99
27DeepSeek V4 Flash 0731max effort36.6%$0.903
28GLM-5.3max effort35.9%$18.38
29MiniMax M3.1 Flashmax effort34.7%$1.26
30GPT-5.6 Lunamax effort30.4%$1.10
31Gemini 3.8 Flashhigh effort30.1%$7.04
32GLM-5.3-Flashmax effort28.4%$1.63
33Gemini 3.7 Flashhigh effort28.3%$5.56
34Grok 4.5xhigh effort27.9%$4.30
34Qwen3.8-27Bmax effort27.9%$6.34
36Step 5 Previewmax effort27.5%$10.22
37Gemini 3.1 Pro Previewhigh effort23.9%$8.13
38Claude Sonnet 4.6max effort23.1%$20.42
39Muse Spark 1.2xhigh effort21.3%$3.74
40Smaug Nemotron 3.5 Nvfp4max effort19.6%$1.08
41Mistral Large 4.0high effort19.2%$1.80
42Muse Spark Openmax effort18.9%$4.24
43Beam 1.0max effort17.3%$9.89
44Nemotron Super VL-3.5max effort15.8%$1.62
45Inkling14.1%$10.54
46Gemma 4 31Bmax effort13.3%$0.814
47MiniMax M3max effort12.4%$4.07
48Mistral Large 4high effort11.8%$9.41
49Nemotron Super VL-3.5 Nvfp4max effort11.5%$1.17
50Claude Haiku 4.5max effort11.1%$2.55
51Nemotron 3.5 Lightning 30B A3Bmax effort8.3%$0.55