BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases9

Claude Opus 5 vs GPT-5.4

Full profiles: Claude Opus 5 (Anthropic) and GPT-5.4 (OpenAI): every benchmark, price and rank.

Updated · Data from 12 sources

As of 25 Oct 2026, Claude Opus 5 (max) leads on quality (68.8 vs 64.7 on the BenchLeader Index), but GPT-5.4 (xhigh) costs 1.8× less, so it is the better buy unless you need the extra quality. Pick Claude Opus 5 for knowledge (72 vs 60); pick GPT-5.4 for long context (66 vs 65).

Claude Opus 5 benchmarks max Anthropic GPT-5.4 benchmarks xhigh OpenAIOpen 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 65 category score; WeirdML 91.8% vs 77.7%Medium
Agents & tool useEither64 vs 64 category score; τ³-Banking (AA) 42.1% vs 39.6%Either
MathsClaude Opus 571 vs 65 category score; FrontierMath Tiers 1–3 85.6% vs 78.6%Medium
ReasoningClaude Opus 574 vs 69 category score; GPQA Diamond 93.9% vs 93.3%Medium
KnowledgeClaude Opus 572 vs 60 category score; SimpleQA Verified 59.9% vs 45.1%High
Long documentsGPT-5.466 vs 65 category score; AA-LCR 82.0% vs 79.3%Medium
Images and other mediaClaude Opus 568 vs 61 category score; MMMU-Pro 84.7% vs 78.4%Medium
Chat quality (human preference)–Only Claude Opus 5 has an Arena rating (69 category score)Not comparable
Lowest costGPT-5.4$5.63 vs $10.00 per 1M blended; agentic coding session $0.210 vs $0.400–
Fastest first answerClaude Opus 524.1 s (Claude Opus 5 (max)) vs 118.0 s (GPT-5.4 (xhigh)) 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 64.7.
  • Claude Opus 5 (max) is stronger in agents & tools, coding, composite, human preference, knowledge, maths, multimodal, reasoning.
  • GPT-5.4 (xhigh) is stronger in long context, instruction following.
  • GPT-5.4 (xhigh) is 1.8× cheaper ($5.63 vs $10.00 per 1M blended).
  • GPT-5.4 (xhigh) streams 1.7× faster (96 vs 57 tokens per second).

At a glance

SpecClaude Opus 5GPT-5.4
DeveloperAnthropicOpenAI
Released24 Jul 20265 Mar 2026
Context window (tokens)1,000,0001,050,000
Compared atClaude Opus 5 (max)GPT-5.4 (xhigh)
Open weightsNoNo
Price in / out / cached input per 1M$5.00 / $25.00 / $0.500$2.50 / $15.00 / $0.250

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 5 (max)GPT-5.4 (xhigh)Leader
Reasoning74.169.1Claude Opus 5
Coding70.264.6Claude Opus 5
Agents & tools64.363.6level
Maths70.665.2Claude Opus 5
Knowledge71.659.7Claude Opus 5
Instruction following–72.3not comparable
Human preference68.5–not comparable
Multimodal67.961.4Claude Opus 5
Long context64.766.1GPT-5.4
Composite86.373.1Claude Opus 5

Where sources disagree

Across the 36 benchmarks both have, Claude Opus 5 leads 31, GPT-5.4 leads 5. Every overall index that rates both puts Claude Opus 5 ahead.

Overall indexClaude Opus 5 (max)GPT-5.4 (xhigh)Leader
BenchLeader Index68.864.7Claude Opus 5
AA Intelligence Index v4.3.250.839.0Claude Opus 5
Arena Text14911475†Claude 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 5 (max)GPT-5.4 (xhigh)
Blended price per 1M tokens$10.00$5.63
Input / output per 1M$5.00 / $25.00$2.50 / $15.00
Cached input per 1M$0.500$0.250
Price per task: chat reply$0.0095$0.0055
Price per task: summarize a 30-page report$0.075$0.039
Price per task: code edit$0.068$0.037
Price per task: agentic coding session$0.400$0.210
Price per task: structured extraction$0.015$0.0080
Cost to run a full benchmark suite$5.86–
Output speed57 tok/s96 tok/s
Time to first answer24.1 s118.0 s
Time to a complete answer32.9 s123.2 s

Each model is compared at its best-ranked reasoning-effort setting. Claude Opus 5 (max) scores 68.8 and takes 24 s to its first answer token; Claude Opus 5 (high) scores 67.9 in 10 s, Claude Opus 5 (xhigh) scores 67.4 in 25 s. GPT-5.4 (xhigh) scores 64.7 and takes 118 s to its first answer token; GPT-5.4 (low) scores 59.7 in 2 s, GPT-5.4 (high) scores 58.9 in 1 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 5 (max)GPT-5.4 (xhigh)
BenchLeader IndexBenchLeader68.864.7
Agents & tools scoreBenchLeader64.363.6
Coding scoreBenchLeader70.264.6
Composite scoreBenchLeader86.373.1
Human preference scoreBenchLeader68.5–
Knowledge scoreBenchLeader71.659.7
Long context scoreBenchLeader64.766.1
Maths scoreBenchLeader70.665.2
Multimodal scoreBenchLeader67.961.4
Reasoning scoreBenchLeader74.169.1
Instruction following scoreBenchLeader–72.3
Blended price $/MBenchLeader$10.00$5.63
Output speedBenchLeader57 tok/s96 tok/s
Time to first answerBenchLeader24.1 s118.0 s
Context windowBenchLeader1,000,000 tokens1,050,000 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub93.9%93.3%
Humanity's Last ExamScale AI / CAIS–36.2%
SimpleBenchSimpleBench80.6%†–
Arena Hard PromptsArena (LMArena)15151487†
LiveBench 2.0 ReasoningLiveBench91.2%†–
LiveBench 1.0 ReasoningLiveBench91.2%88.1%
GPQA Diamond (AA)Artificial Analysis93.2%92.0%
Humanity's Last Exam (AA)Artificial Analysis54.9%43.7%
GPQA Diamond (Vals)Vals AI93.4%†91.7%
Kagi LLM BenchmarkKagi LLM Benchmark–63.8%†
ARC-AGI-1ARC Prize97.5%93.7%
ARC-AGI-2ARC Prize90.4%74.0%
ARC-AGI-3ARC Prize30.2%†0.2%†
CritPtArtificial Analysis29.1%23.4%
MysteryMechanismVals AI37.4%†–
Chess PuzzlesEpoch AI Benchmarking Hub43.0%48.0%
Mystery Game PuzzlesEpoch AI Benchmarking Hub59.0%37.0%
CL-benchEpoch AI Benchmarking Hub–27.9%
CL-bench LifeEpoch AI Benchmarking Hub–21.7%
LMCAEpoch AI Benchmarking Hub63.3%52.0%
DTBenchEpoch AI Benchmarking Hub97.6%94.4%
ForecastBenchForecastBench66.5%63.1%†
BTF-3Epoch AI Benchmarking Hub12.0%†–
Coding
SWE-bench Verified (Epoch)Epoch AI Benchmarking Hub–76.9%†
SciCodeSciCode56.4%56.6%
Show all 162 benchmarks
MetricPublisherClaude Opus 5 (max)GPT-5.4 (xhigh)
Coding
WeirdMLWeirdML91.8%77.7%
FrontierCodeCognition53.4%–
GSO-BenchGSO-Bench–31.4%
Arena CodingArena (LMArena)15281511†
Arena WebDevArena (LMArena)16911401†
LiveBench 2.0 CodingLiveBench90.8%†–
LiveBench 1.0 CodingLiveBench81.5%77.5%
SciCode (AA)Artificial Analysis56.4%–
LiveCodeBenchVals AI89.0%†84.1%
IOIVals AI84.3%†–
SWE-bench (Vals)Vals AI97.0%†78.2%
SWE-Bench ProScale AI SEAL–59.1%
Code MigrationVals AI57.5%†35.0%
ProgramBenchVals AI3.0%†0.0%
Vibe Code Bench 1-100Vals AI28.5%†–
Vibe Code Bench v1.1Vals AI88.4%†67.4%
SWE-Bench Pro (private)Scale AI SEAL–59.1%
SWE Atlas: Codebase QnAScale AI SEAL63.2%†40.8%
SWE Atlas: RefactoringScale AI SEAL–44.3%
SWE Atlas: Test WritingScale AI SEAL62.2%†44.4%
MirrorCodeEpoch AI Benchmarking Hub–15.6%†
DeepSWE v1.1Epoch AI Benchmarking Hub73.7%51.8%
CursorBenchCursor46.6%–
ALE-BenchEpoch AI Benchmarking Hub2164.6†1607†
AlgoTuneAlgoTune–1.9†
FrontierSWEEpoch AI Benchmarking Hub52.0%–
Terminal-Bench 4.0 (AA)Artificial Analysis49.0%–
Terminal-Bench 2.1 (AA)Artificial Analysis89.1%78.3%
Startrise FrontendStartrise82.3%†–
Agents & tools
Terminal-BenchTerminal-Bench51.8%81.8%†
OSWorld-Verified 2.0OSWorld31.4%–
APEX-AgentsMercor65.8%52.4%†
Arena AgentArena (LMArena)8.2–
LiveBench Long AgenticLiveBench62.1%†–
LiveBench BrowserLiveBench84.0%†–
LiveBench Agentic CodingLiveBench65.2%53.8%
Terminal-Bench HardArtificial Analysis–57.6%
τ²-Bench Telecom (AA)Artificial Analysis–87.1%
Terminal-Bench 2.1 (Vals)Vals AI84.6%†–
MCP AtlasScale AI SEAL85.8%†70.6%
HiL-BenchScale AI SEAL57.0%†9.7%†
τ³-bench Bankingτ²-bench48.7%39.4%
τ³-Banking (AA)Artificial Analysis42.1%39.6%
Analyst Agent (AA)Artificial Analysis53.8%–
APEX-Agents (AA)Artificial Analysis–33.3%
CUA-benchVals AI9.0%–
CyberBenchVals AI65.4%†–
Finance Agent v2Vals AI58.6%†–
Harvey's Legal Agent Benchmark v1Vals AI6.7%†0.0%
Legal Research BenchVals AI55.3%†–
SkillsBenchVals AI60.4%†51.7%
SREBenchVals AI12.2%†–
Tax Agent BenchVals AI45.5%†–
Terminal-Bench 2.0 (Vals)Vals AI–58.4%†
Terminal-Bench 4.0 (Vals)Vals AI53.5%†–
Terminal-Bench ScienceVals AI27.1%†–
Time Horizon Index: KSPVals AI18.8%†–
Arena SearchArena (LMArena)1214†–
BALROGBALROG63.4%–
DeepResearch BenchDeepResearch Bench–35.1%†
PostTrainBenchEpoch AI Benchmarking Hub35.0%†19.0%†
METR Time HorizonsMETR–74.3%
GBAEvalEpoch AI Benchmarking Hub79.6%†45.1%†
Vending-Bench 2Andon Labs11181.9†6144.2†
AutomationBenchArtificial Analysis56.6%–
GDP.pdf (AA)Artificial Analysis21.6%–
EnterpriseOps-GymArtificial Analysis47.5%–
AA-Briefcase v1.1Artificial Analysis1657–
Terminal-Bench-Science 0.1 (AA)Artificial Analysis28.6%–
DeepSec-Bench (AA)Artificial Analysis25.9%–
DeepSearchQAGoogle–63.7%†
FACTS SearchGoogle85.3%†–
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub85.6%78.6%
FrontierMath Tier 4Epoch AI Benchmarking Hub73.2%49.0%
OTIS Mock AIMEEpoch AI Benchmarking Hub98.9%95.3%
ProofBenchVals AI99.0%56.0%
LiveBench MathematicsLiveBench95.7%94.2%
AIME (Vals)Vals AI–96.7%
AIME 2026MathArena–99.2%
HMMT February 2026MathArena–97.7%
USAMO 2026MathArena–95.2%
MathArena ApexMathArena–54.2%
Arena MathsArena (LMArena)15171460†
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub59.9%45.1%
LiveBench 2.0 Data AnalysisLiveBench74.6%†–
LiveBench ResearchLiveBench88.9%†–
LiveBench 1.0 Data AnalysisLiveBench74.5%79.3%
AA-OmniscienceArtificial Analysis37.15.8
MMLU-ProVals AI91.6%†87.5%
LegalBenchVals AI87.0%†86.0%
CorpFinVals AI73.2%†65.3%
TaxEvalVals AI75.1%†74.0%
MedQAVals AI–96.1%
PRBench FinanceScale AI SEAL–45.6%†
PRBench LegalScale AI SEAL–44.4%†
MultiNRCScale AI SEAL–58.3%†
BioMysteryBenchVals AI79.3%†–
CaseLaw v2Vals AI–63.8%
Excel Modeling BenchmarkVals AI73.6%†–
MedCodeVals AI63.6%†41.3%
MedScribeVals AI91.0%†77.5%
Public Benefits Bench v1.1Vals AI76.9%†–
DrugDiscoveryBenchScale AI SEAL53.7%–
EBR-benchEpoch AI Benchmarking Hub45.7%25.4%
GDP.pdfEpoch AI Benchmarking Hub24.0%–
AA-Omniscience: accuracyArtificial Analysis60.9%50.9%
AA-Omniscience: non-hallucinationArtificial Analysis39.2%8.3%
Instruction following
LiveBench LanguageLiveBench88.7%82.6%
LiveBench Instruction FollowingLiveBench63.8%70.2%
IFBenchArtificial Analysis–74.0%
Arena Instruction FollowingArena (LMArena)14951460†
Human preference
Arena TextArena (LMArena)14911465†
EQ-Bench 4EQ-Bench1360†1260†
Arena Creative WritingArena (LMArena)14741435†
Arena Multi-turnArena (LMArena)14871480†
Arena Longer QueriesArena (LMArena)15031477†
Multimodal
Arena VisionArena (LMArena)1320†1292†
LiveBench DesignLiveBench83.0%†–
MMMU-ProArtificial Analysis84.7%78.4%
VISTAScale AI SEAL–50.9%†
MMMU-Pro (Vals)Vals AI89.9%†87.5%
MortgageTaxVals AI72.1%†68.3%
SAGEVals AI49.4%†43.3%
VTBScale AI SEAL–29.2%†
Blueprint-Bench 2Epoch AI Benchmarking Hub30.4%†27.1%†
Long context
AA-LCRArtificial Analysis79.3%82.0%
Arena DocumentArena (LMArena)1490†1470†
MLCRArtificial Analysis11.1%–
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub163.9†157.3†
LiveBench 2.0LiveBench82.1%†–
LiveBench 1.0LiveBench80.1%78.0%
AA Intelligence Index v4.3.2Artificial Analysis50.839.0
Vals RSI IndexVals AI33.0%15.5%
Vals Science IndexVals AI47.9%†–
Vals Multimodal IndexVals AI73.9%†–
Safety & honesty
FORTRESSScale AI SEAL16.8%†–
MASKScale AI SEAL–89.7%†

Data as of 2026-10-25. Best configuration of each model; every score links to its source on the model pages.

What changed

  • Claude Opus 5 (max): 1 new result (Vals RSI Index).
  • GPT-5.4 (xhigh): 2 new results (Harvey's Legal Agent Benchmark v1, 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…).
  • GPT-5.4 (xhigh): 2 new results (USAMO 2026, τ³-bench Banking).
  • Claude Opus 5 (max): 1 score revision.
  • Claude Opus 5 (max): 2 score revisions.
  • GPT-5.4 (xhigh): 1 score revision.

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

Claude Opus 5 vs GPT-5.4: questions

Is Claude Opus 5 better than GPT-5.4?
Claude Opus 5 (max) leads on quality: 68.8 vs 64.7. The BenchLeader Index combines every independent quality benchmark; Claude Opus 5 (max) is ahead overall as of 2026-10-25, but check the category scores for your use.
Which is cheaper, Claude Opus 5 or GPT-5.4, and will that change?
GPT-5.4 is cheaper today: $5.63 against $10.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Is Claude Opus 5 better than GPT-5.4 for coding?
Claude Opus 5 scores higher in coding (70 vs 65 on the category index, where 50 is average). On SciCode, the top shared coding benchmark, it is 56.4% for Claude Opus 5 against 56.6% for GPT-5.4.
Is Claude Opus 5 better than GPT-5.4 for agents and tool use?
They are level in agentic tasks (64 vs 64 on the category index). On τ³-Banking (AA), the top shared agents & tools benchmark, it is 42.1% for Claude Opus 5 against 39.6% for GPT-5.4.
Which is better for long documents?
GPT-5.4 scores higher on long-context benchmarks (66 vs 65). Both take about 1.1M tokens: Claude Opus 5 1,000,000, GPT-5.4 1,050,000.
Which has lower latency?
Claude Opus 5 (max) answers first: 24.1 s against 118.0 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 5 or GPT-5.4?
GPT-5.4 streams faster: 96 against 57 output tokens per second.
Do independent benchmarks agree on Claude Opus 5 vs GPT-5.4?
Across the 36 benchmarks both have, Claude Opus 5 leads 31, GPT-5.4 leads 5. 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.