head to head
Claude Haiku 4.5 vs DeepSeek V4 Pro
1 shared benchmarks
0
wins
0
ties
1
wins
73.3%
SWE-bench Verified
80.6%
benchmark.space canonical public ledger of AI capability