Claude Fable 5 vs Kimi K3
25 shared benchmarks
14
wins
1
ties
10
wins
43.3%
APEX-Agents
41.0%
29.1%
AutomationBench
30.8%
90.5%
BabyVision
85.7%
88.0%
BrowseComp
90.4%
88.9%
CharXiv
84.8%
70.0%
DeepSWE
67.3%
86.6%
FrontierSWE
81.2%
92.6%
GPQA Diamond
93.5%
53.3%
Humanity's Last Exam
43.5%
57.4%
JobBench
52.9%
94.8%
MATH-Vision
94.3%
84.7%
MCP-Atlas
84.2%
49.9%
MLS Bench
48.3%
81.2%
MMMU Pro
81.6%
69.9%
OfficeQA Pro
63.3%
89.8%
OmniDocBench
91.1%
57.2%
PerceptionBench
58.5%
41.4%
PostTrain Bench
36.6%
76.8%
Program Bench
77.8%
35.0%
SWE Marathon
42.0%
34.7%
SpreadsheetBench 2
34.8%
84.6%
Terminal-Bench 2.1
88.3%
77.9%
Toolathlon-Verified
73.2%
56.7%
WorldVQA
51.0%
23.0%
ZeroBench
23.0%