Compare / Nova 2 Lite vs Claude Opus 4.5

Nova 2 Lite vs Claude Opus 4.5

20 results from 4 sources that measured both models. Each row is in the source's own units; there is no overall winner.

Nova 2 Lite
Amazon · profile
Claude Opus 4.5
Anthropic · profile
Results compared
20
Shared sources
4
BenchmarkMetricNova 2 LiteClaude Opus 4.5
Arena (formerly LMArena)reportedOverallArena rating 1,3801,459
Arena (formerly LMArena)reportedBusiness, management and financeArena rating 1,3511,472
Arena (formerly LMArena)reportedCreative writingArena rating 1,2691,469
Arena (formerly LMArena)reportedExpert promptsArena rating 1,3541,503
Arena (formerly LMArena)reportedInstruction followingArena rating 1,3271,483
Arena (formerly LMArena)reportedOverallArena rating 1,3351,474
Arena (formerly LMArena)reportedWriting, literature and languageArena rating 1,2981,467
Berkeley Function Calling Leaderboard (BFCL) V4reportedIrrelevance detection% correct 82.1%90.8%
Berkeley Function Calling Leaderboard (BFCL) V4reportedMemory% correct 2.4%73.8%
Berkeley Function Calling Leaderboard (BFCL) V4reportedMulti-turn tasks% correct 2.1%68.4%
Berkeley Function Calling Leaderboard (BFCL) V4reportedOverall accuracy% correct 27.1%77.5%
Berkeley Function Calling Leaderboard (BFCL) V4reportedRelevance detection% correct 75.0%68.8%
Berkeley Function Calling Leaderboard (BFCL) V4reportedSingle-turn calls (curated)% correct 87.0%89.7%
Berkeley Function Calling Leaderboard (BFCL) V4reportedSingle-turn calls (user-contributed)% correct 80.8%79.8%
Berkeley Function Calling Leaderboard (BFCL) V4reportedWeb search% correct 5.0%84.5%
UGI LeaderboardreportedRequested-length error% off the requested word count 45.0%3.0%
UGI LeaderboardreportedStyle adherencescore from 0 to 1 0.350.39
UGI LeaderboardreportedWriting scorescore out of 100 23.570.4
Vectara Hallucination LeaderboardreportedAnswer rate% of documents 99.6%98.7%
Vectara Hallucination LeaderboardreportedHallucination rate% of summaries 5.1%10.9%

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.