Compare / Claude Opus 4.5 vs Claude Opus 4.8

Claude Opus 4.5 vs Claude Opus 4.8

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

Claude Opus 4.5
Anthropic · profile
Claude Opus 4.8
Anthropic · profile
Results compared
20
Shared sources
4
BenchmarkMetricClaude Opus 4.5Claude Opus 4.8
Arena (formerly LMArena)reportedOverallArena rating 1,4651,486
Arena (formerly LMArena)reportedOverallArena rating 1,1871,195
Arena (formerly LMArena)reportedOverallArena rating 1,4591,469
Arena (formerly LMArena)reportedBusiness, management and financeArena rating 1,4721,490
Arena (formerly LMArena)reportedCreative writingArena rating 1,4691,467
Arena (formerly LMArena)reportedExpert promptsArena rating 1,5031,525
Arena (formerly LMArena)reportedInstruction followingArena rating 1,4831,491
Arena (formerly LMArena)reportedOverallArena rating 1,4741,480
Arena (formerly LMArena)reportedWriting, literature and languageArena rating 1,4671,474
OpenHands IndexreportedAverage% resolved 60.6%71.9%
OpenHands IndexreportedFront end (SWE-Bench Multimodal)% resolved 41.2%50.0%
OpenHands IndexreportedGreenfield (Commit0)% resolved 37.5%62.5%
OpenHands IndexreportedInformation gathering (GAIA)% resolved 69.1%78.8%
OpenHands IndexreportedIssue resolution (SWE-Bench)% resolved 76.6%83.8%
OpenHands IndexreportedTesting (SWT-Bench)% resolved 78.5%84.3%
tau2-benchreportedConsistency (pass^4)% of tasks 11.3%22.7%
tau2-benchreportedTask success (pass^1)% of tasks 24.7%39.7%
UGI LeaderboardreportedRequested-length error% off the requested word count 3.0%6.0%
UGI LeaderboardreportedStyle adherencescore from 0 to 1 0.390.40
UGI LeaderboardreportedWriting scorescore out of 100 70.465.9

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.