Compare / Qwen3.6 Plus vs Claude Haiku 4.5

Qwen3.6 Plus vs Claude Haiku 4.5

22 results from 5 sources that measured both models. Each row is in the source's own units; there is no overall winner.

Qwen3.6 Plus
Alibaba · profile
Claude Haiku 4.5
Anthropic · profile
Results compared
22
Shared sources
5
BenchmarkMetricQwen3.6 PlusClaude Haiku 4.5
Arena (formerly LMArena)reportedOverallArena rating 1,4381,418
Arena (formerly LMArena)reportedBusiness, management and financeArena rating 1,4461,420
Arena (formerly LMArena)reportedCreative writingArena rating 1,4071,387
Arena (formerly LMArena)reportedExpert promptsArena rating 1,4761,452
Arena (formerly LMArena)reportedInstruction followingArena rating 1,4361,415
Arena (formerly LMArena)reportedOverallArena rating 1,4431,414
Arena (formerly LMArena)reportedWriting, literature and languageArena rating 1,4231,399
Artificial AnalysisreportedArtificial Analysis Coding Indexindex score 54.543.9
Artificial AnalysisreportedArtificial Analysis Intelligence Indexindex score 27.016.9
Artificial AnalysisreportedGPQA Diamond% of questions 88.2%67.2%
Artificial AnalysisreportedHumanity's Last Exam% of questions 27.8%10.4%
Artificial AnalysisreportedIFBench% of instructions 75.2%54.3%
Artificial AnalysisreportedLong-context reasoning (AA-LCR)% of questions 78.3%74.3%
Artificial AnalysisreportedTerminal-Bench 2.1% of tasks 61.4%44.2%
Artificial AnalysisreportedTerminal-Bench Hard% of tasks 43.9%27.3%
Artificial AnalysisreportedΤ-bench banking% of tasks 20.8%9.3%
Artificial AnalysisreportedΤ²-bench telecom% of tasks 97.7%54.7%
SimpleQA Verified (Epoch AI)reportedCorrect answers% of questions 44.1%13.2%
UGI LeaderboardreportedRequested-length error% off the requested word count 8.0%6.0%
UGI LeaderboardreportedStyle adherencescore from 0 to 1 0.350.39
UGI LeaderboardreportedWriting scorescore out of 100 58.252.2
Vending-Bench 2reportedMoney after a yearUS dollars $5,114.87$458.89

Bold green marks the better value on that metric. Where a source reports ranges that overlap, the difference may not be meaningful; see the benchmark page for ranges.