BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases9

Claude Opus 4.5 vs Claude Opus 5

Full profiles: Claude Opus 4.5 (Anthropic) and Claude Opus 5 (Anthropic): every benchmark, price and rank.

Updated · Data from 13 sources

As of 26 Oct 2026, Claude Opus 5 (max) is the better pick for most workloads: it leads on quality (68.8 vs 59.9 on the BenchLeader Index) at about the same price. Pick Claude Opus 4.5 for agents & tools (75 vs 64); pick Claude Opus 5 for reasoning (74 vs 54).

Claude Opus 4.5 benchmarks thinking Anthropic Claude Opus 5 benchmarks max AnthropicOpen in the comparison tool

Which to pick for…

The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.

Use casePickWhyConfidence
CodingClaude Opus 570 vs 61 category scoreHigh
Agents & tool useClaude Opus 4.575 vs 64 category scoreHigh
MathsClaude Opus 571 vs 64 category scoreMedium
ReasoningClaude Opus 574 vs 54 category score; ARC-AGI-2 90.4% vs 37.6%High
KnowledgeClaude Opus 572 vs 60 category score; AA-Omniscience 37.1 vs 14High
Long documentsClaude Opus 565 vs 64 category score; AA-LCR 79.3% vs 77.3%Medium
Images and other mediaClaude Opus 568 vs 57 category score; MMMU-Pro 84.7% vs 74.0%High
Chat quality (human preference)–Only Claude Opus 5 has an Arena rating (69 category score)Not comparable
Lowest costEither$10.00 vs $10.00 per 1M blended; agentic coding session $0.400 vs $0.400–
Fastest first answerClaude Opus 4.513.8 s (Claude Opus 4.5 (thinking)) vs 68.7 s (Claude Opus 5 (max)) to the first answer token–
High volume on a budgetNeither flagshipA smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs–

Key differences

  • Claude Opus 5 (max) leads on quality: 68.8 vs 59.9.
  • Claude Opus 4.5 (thinking) is stronger in agents & tools, instruction following.
  • Claude Opus 5 (max) is stronger in coding, composite, knowledge, long context, maths, multimodal, reasoning, human preference.
  • They cost about the same ($10.00 per 1M blended).

At a glance

SpecClaude Opus 4.5Claude Opus 5
DeveloperAnthropicAnthropic
Released24 Nov 202524 Jul 2026
Context window (tokens)200,0001,000,000
Compared atClaude Opus 4.5 (thinking)Claude Opus 5 (max)
Open weightsNoNo
Price in / out / cached input per 1M$5.00 / $25.00 / $0.500$5.00 / $25.00 / $0.500

Category by category

Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.

CategoryClaude Opus 4.5 (thinking)Claude Opus 5 (max)Leader
Reasoning54.474.1Claude Opus 5
Coding60.670.2Claude Opus 5
Agents & tools74.864.3Claude Opus 4.5
Maths64.070.6Claude Opus 5
Knowledge59.971.6Claude Opus 5
Instruction following53.5–not comparable
Human preference–68.5not comparable
Multimodal57.367.9Claude Opus 5
Long context63.764.7Claude Opus 5
Composite62.086.3Claude Opus 5

Where sources disagree

Across the 10 benchmarks both have, Claude Opus 4.5 leads 0, Claude Opus 5 leads 10. Every overall index that rates both puts Claude Opus 5 ahead.

Overall indexClaude Opus 4.5 (thinking)Claude Opus 5 (max)Leader
BenchLeader Index59.968.8Claude Opus 5
AA Intelligence Index v4.3.229.150.8Claude Opus 5
Arena Text1474†1491Claude Opus 5

Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).

Price, speed and latency

List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page; suite cost is what Artificial Analysis paid to run its whole Intelligence Index.

MeasureClaude Opus 4.5 (thinking)Claude Opus 5 (max)
Blended price per 1M tokens$10.00$10.00
Input / output per 1M$5.00 / $25.00$5.00 / $25.00
Cached input per 1M$0.500$0.500
Price per task: chat reply$0.0095$0.0095
Price per task: summarize a 30-page report$0.075$0.075
Price per task: code edit$0.068$0.068
Price per task: agentic coding session$0.400$0.400
Price per task: structured extraction$0.015$0.015
Cost to run a full benchmark suite–$5.86
Output speed55 tok/s51 tok/s
Time to first answer13.8 s68.7 s
Time to a complete answer22.9 s78.4 s

Each model is compared at its best-ranked reasoning-effort setting. Claude Opus 4.5 (thinking) scores 59.9 and takes 14 s to its first answer token; Claude Opus 4.5 (high) scores 59.2 in 1 s, Claude Opus 4.5 scores 57.9 in 1 s. Claude Opus 5 (max) scores 68.8 and takes 69 s to its first answer token; Claude Opus 5 (high) scores 67.9 in 17 s, Claude Opus 5 (xhigh) scores 67.4 in 31 s. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.

Every benchmark

162 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.

MetricPublisherClaude Opus 4.5 (thinking)Claude Opus 5 (max)
BenchLeader IndexBenchLeader59.968.8
Agents & tools scoreBenchLeader74.864.3
Coding scoreBenchLeader60.670.2
Composite scoreBenchLeader62.086.3
Instruction following scoreBenchLeader53.5–
Knowledge scoreBenchLeader59.971.6
Long context scoreBenchLeader63.764.7
Maths scoreBenchLeader64.070.6
Multimodal scoreBenchLeader57.367.9
Reasoning scoreBenchLeader54.474.1
Human preference scoreBenchLeader–68.5
Blended price $/MBenchLeader$10.00$10.00
Output speedBenchLeader55 tok/s51 tok/s
Time to first answerBenchLeader13.8 s68.7 s
Context windowBenchLeader200,000 tokens1,000,000 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub86.0%†93.9%
Humanity's Last ExamScale AI / CAIS25.2%†–
SimpleBenchSimpleBench62.0%†80.6%†
Arena Hard PromptsArena (LMArena)1498†1515
LiveBench 1.0 ReasoningLiveBench80.1%†91.2%
GPQA Diamond (AA)Artificial Analysis86.6%93.2%
Humanity's Last Exam (AA)Artificial Analysis30.1%54.9%
GPQA Diamond (Vals)Vals AI85.9%93.4%†
Kagi LLM BenchmarkKagi LLM Benchmark80.2%–
ARC-AGI-1ARC Prize80.0%97.5%
ARC-AGI-2ARC Prize37.6%90.4%
ARC-AGI-3ARC Prize–30.2%†
CritPtArtificial Analysis4.6%29.1%
MysteryMechanismVals AI–37.4%†
SciPredictScale AI SEAL23.1%†–
Chess PuzzlesEpoch AI Benchmarking Hub14.0%†43.0%
Mystery Game PuzzlesEpoch AI Benchmarking Hub22.0%†59.0%
CL-benchEpoch AI Benchmarking Hub21.1%†–
LMCAEpoch AI Benchmarking Hub44.5%†63.3%
DTBenchEpoch AI Benchmarking Hub89.9%†97.6%
ForecastBenchForecastBench64.9%†66.5%
BTF-3Epoch AI Benchmarking Hub–12.0%†
Coding
SWE-bench Verified (Epoch)Epoch AI Benchmarking Hub76.7%†–
SciCodeSciCode–56.4%
WeirdMLWeirdML63.7%†91.8%
Show all 162 benchmarks
MetricPublisherClaude Opus 4.5 (thinking)Claude Opus 5 (max)
Coding
FrontierCodeCognition–53.4%
GSO-BenchGSO-Bench26.5%†–
Arena CodingArena (LMArena)1523†1528
Arena WebDevArena (LMArena)1470†1691
LiveBench 1.0 CodingLiveBench79.7%†81.5%
LiveCodeBenchVals AI83.7%89.0%†
IOIVals AI–84.3%†
SWE-bench (Vals)Vals AI76.4%97.0%†
SWE-Bench ProScale AI SEAL45.9%†–
SWE-bench Verified (bash only)SWE-bench76.8%†–
SWE-bench Verified (any scaffold)SWE-bench79.2%†–
Code MigrationVals AI–57.5%†
ProgramBenchVals AI–3.0%†
Vibe Code Bench 1-100Vals AI–28.5%†
Vibe Code Bench v1.1Vals AI20.6%88.4%†
SWE-Bench Pro (private)Scale AI SEAL45.9%†–
SWE Atlas: Codebase QnAScale AI SEAL–63.2%†
SWE Atlas: Test WritingScale AI SEAL–62.2%†
DeepSWE v1.1Epoch AI Benchmarking Hub–73.7%
CursorBenchCursor–46.6%
ALE-BenchEpoch AI Benchmarking Hub1025.4†2164.6†
AlgoTuneAlgoTune1.8†–
FrontierSWEEpoch AI Benchmarking Hub–52.0%
Terminal-Bench 4.0 (AA)Artificial Analysis–49.0%
Terminal-Bench 2.1 (AA)Artificial Analysis–89.1%
Startrise FrontendStartrise–82.3%†
Agents & tools
Terminal-BenchTerminal-Bench63.1%†51.8%
OSWorld-Verified 2.0OSWorld–31.4%
GDPvalOpenAI45.5%†–
CybenchCybench82.0%†–
Remote Labor IndexScale AI / CAIS3.8%†–
APEX-AgentsMercor–65.8%
Arena AgentArena (LMArena)–8.2
LiveBench Agentic CodingLiveBench39.7%†65.2%
Terminal-Bench HardArtificial Analysis47.0%–
τ²-Bench Telecom (AA)Artificial Analysis89.5%–
Terminal-Bench 2.1 (Vals)Vals AI–84.6%†
MCP AtlasScale AI SEAL69.8%†85.8%†
HiL-BenchScale AI SEAL–57.0%†
BFCL OverallBerkeley Function Calling Leaderboard77.5%†–
τ²-benchτ²-bench85.3%†–
τ³-bench Bankingτ²-bench24.7%†48.7%
τ³-Banking (AA)Artificial Analysis–42.1%
Analyst Agent (AA)Artificial Analysis–53.8%
CUA-benchVals AI–9.0%
CyberBenchVals AI–65.4%†
Finance Agent v2Vals AI–58.6%†
Harvey's Legal Agent Benchmark v1Vals AI–6.7%†
Legal Research BenchVals AI–55.3%†
Poker AgentVals AI1033.4%†–
SkillsBenchVals AI–60.4%†
SREBenchVals AI–12.2%†
Tax Agent BenchVals AI–45.5%†
Terminal-Bench 2.0 (Vals)Vals AI53.9%–
Terminal-Bench 4.0 (Vals)Vals AI–53.5%†
Terminal-Bench ScienceVals AI–27.1%†
Time Horizon Index: KSPVals AI–18.8%†
Arena SearchArena (LMArena)–1214†
BALROGBALROG43.0%†63.4%
DeepResearch BenchDeepResearch Bench54.8%†–
PostTrainBenchEpoch AI Benchmarking Hub–35.0%†
METR Time HorizonsMETR75.0%†–
GBAEvalEpoch AI Benchmarking Hub–79.6%†
Vending-Bench 2Andon Labs4967.1†11181.9†
AutomationBenchArtificial Analysis–56.6%
GDP.pdf (AA)Artificial Analysis–21.6%
EnterpriseOps-GymArtificial Analysis–47.5%
AA-Briefcase v1.1Artificial Analysis–1657
Terminal-Bench-Science 0.1 (AA)Artificial Analysis–28.6%
DeepSec-Bench (AA)Artificial Analysis–25.9%
DeepSearchQAGoogle24.0%–
FACTS SearchGoogle–85.3%†
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub34.4%†85.6%
FrontierMath Tier 4Epoch AI Benchmarking Hub4.9%†73.2%
OTIS Mock AIMEEpoch AI Benchmarking Hub86.1%†98.9%
ProofBenchVals AI36.0%†99.0%
LiveBench MathematicsLiveBench90.4%†95.7%
AIME (Vals)Vals AI95.4%–
MGSMVals AI95.2%–
Arena MathsArena (LMArena)1463†1517
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub45.7%†59.9%
LiveBench 1.0 Data AnalysisLiveBench74.4%†74.5%
AA-OmniscienceArtificial Analysis1437.1
MMLU-ProVals AI87.3%91.6%†
LegalBenchVals AI84.6%87.0%†
CorpFinVals AI65.1%73.2%†
TaxEvalVals AI74.9%75.1%†
MedQAVals AI95.9%–
PRBench FinanceScale AI SEAL46.2%–
PRBench LegalScale AI SEAL44.2%–
MultiNRCScale AI SEAL48.6%–
BioMysteryBenchVals AI–79.3%†
CaseLaw v2Vals AI62.6%–
Excel Modeling BenchmarkVals AI–73.6%†
MedCodeVals AI49.2%63.6%†
MedScribeVals AI85.3%91.0%†
Public Benefits Bench v1.1Vals AI–76.9%†
DrugDiscoveryBenchScale AI SEAL–53.7%
EBR-benchEpoch AI Benchmarking Hub14.3%†45.7%
GDP.pdfEpoch AI Benchmarking Hub–24.0%
AA-Omniscience: accuracyArtificial Analysis46.6%60.9%
AA-Omniscience: non-hallucinationArtificial Analysis39.0%39.2%
Global-MMLU-Lite (AA)Artificial Analysis91.3%–
Instruction following
LiveBench LanguageLiveBench81.3%†88.7%
LiveBench Instruction FollowingLiveBench62.5%†63.8%
IFBenchArtificial Analysis58.0%–
MultiChallengeScale AI SEAL59.0%–
TutorBenchScale AI SEAL51.2%–
Arena Instruction FollowingArena (LMArena)1474†1495
Human preference
Arena TextArena (LMArena)1470†1491
EQ-Bench 4EQ-Bench–1360†
Arena Creative WritingArena (LMArena)1462†1474
Arena Multi-turnArena (LMArena)1484†1487
Arena Longer QueriesArena (LMArena)1492†1503
Multimodal
Arena VisionArena (LMArena)–1320†
MMMU-ProArtificial Analysis74.0%84.7%
VISTAScale AI SEAL46.4%–
MMMU (validation)MMMU80.7%†–
MMMU-Pro (official)MMMU73.9%†–
MMMU-Pro (Vals)Vals AI83.0%89.9%†
MortgageTaxVals AI67.7%72.1%†
SAGEVals AI52.1%49.4%†
GeoBenchGeoBench75.0%†–
VPCTEpoch AI Benchmarking Hub40.0%†–
Blueprint-Bench 2Epoch AI Benchmarking Hub–30.4%†
Long context
AA-LCRArtificial Analysis77.3%79.3%
Arena DocumentArena (LMArena)1465†1490†
MLCRArtificial Analysis–11.1%
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub150.1†163.9†
LiveBench 1.0LiveBench72.6%†80.1%
AA Intelligence Index v4.3.2Artificial Analysis29.150.8
Vals RSI IndexVals AI14.8%†33.0%
Vals Science IndexVals AI–47.9%†
Vals Multimodal IndexVals AI–73.9%†
Safety & honesty
FORTRESSScale AI SEAL9.6%16.8%†
MASKScale AI SEAL92.5%–
Metadata
AA Openness IndexArtificial Analysis11.1–

Data as of 2026-10-26. Best configuration of each model; every score links to its source on the model pages.

What changed

  • Claude Opus 4.5 (thinking): 1 new result (Global-MMLU-Lite (AA)).
  • Claude Opus 5 (max): 1 new result (Vals RSI Index).
  • Claude Opus 5 (max): 1 score revision.
  • Claude Opus 5 (max): 1 new result (ForecastBench); 1 score revision.
  • Claude Opus 5 (max): BenchLeader Index 68.3 → 68.9.
  • Claude Opus 5 (max): 4 new results (GDPval-AA v2.1, AA-Omniscience: accuracy, AA-Omniscience: non-hallucination…).
  • Claude Opus 5 (max): 1 score revision.
  • Claude Opus 5 (max): 2 score revisions.
  • Claude Opus 4.5 (thinking): 1 new result (PRBench Legal).
  • Claude Opus 5 (max): 2 new results (Terminal-Bench-Science 0.1 (AA), DeepSec-Bench (AA)).

Rankings: Best LLM overall · Best LLM for coding · Best LLM for agents · Best cheap LLMs

Claude Opus 4.5 vs Claude Opus 5: questions

Is Claude Opus 4.5 better than Claude Opus 5?
Claude Opus 5 (max) leads on quality: 68.8 vs 59.9. The BenchLeader Index combines every independent quality benchmark; Claude Opus 5 (max) is ahead overall as of 2026-10-26, but check the category scores for your use.
Which is cheaper, Claude Opus 4.5 or Claude Opus 5, and will that change?
They cost the same: $10.00 per million tokens blended. No price change has been announced for either.
Is Claude Opus 4.5 better than Claude Opus 5 for coding?
Claude Opus 5 scores higher in coding (70 vs 61 on the category index, where 50 is average).
Is Claude Opus 4.5 better than Claude Opus 5 for agents and tool use?
Claude Opus 4.5 scores higher in agentic tasks (75 vs 64 on the category index, where 50 is average).
Which is better for long documents?
Claude Opus 5 scores higher on long-context benchmarks (65 vs 64). Claude Opus 5 accepts more context: 1,000,000 against 200,000 tokens for Claude Opus 4.5.
Which has lower latency?
Claude Opus 4.5 (thinking) answers first: 13.8 s against 68.7 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
Which is faster, Claude Opus 4.5 or Claude Opus 5?
Claude Opus 4.5 streams faster: 55 against 51 output tokens per second.
Do independent benchmarks agree on Claude Opus 4.5 vs Claude Opus 5?
Across the 10 benchmarks both have, Claude Opus 4.5 leads 0, Claude Opus 5 leads 10. Every overall index that rates both puts Claude Opus 5 ahead.
When was Claude Opus 5 released?
Claude Opus 5 was released on 24 Jul 2026.