BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases9

Claude Opus 4.8 vs GPT-5.3-Codex

Full profiles: Claude Opus 4.8 (Anthropic) and GPT-5.3-Codex (OpenAI): every benchmark, price and rank.

Updated · Data from 11 sources

As of 26 Oct 2026, Claude Opus 4.8 (max) and GPT-5.3-Codex (xhigh) are neck and neck on quality (63.3 vs 63.8 on the BenchLeader Index, inside the ±6.9 margin of error), and GPT-5.3-Codex (xhigh) costs 2.1× less, which makes it the better buy for most workloads. GPT-5.3-Codex leads in every category we measure both in.

Claude Opus 4.8 benchmarks max Anthropic GPT-5.3-Codex 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
CodingGPT-5.3-Codex61 vs 59 category scoreMedium
Agents & tool useGPT-5.3-Codex80 vs 57 category score; Terminal-Bench Hard 53.0% vs 58.3%High
ReasoningEither69 vs 69 category score; CritPt 20.9% vs 16.9%Either
KnowledgeEither67 vs 66 category score; AA-Omniscience 28.8 vs 10.9Either
Long documentsGPT-5.3-Codex67 vs 64 category score; AA-LCR 83.3% vs 77.7%Medium
Following instructionsGPT-5.3-Codex74 vs 62 category score; IFBench 75.4% vs 62.2%High
Lowest costGPT-5.3-Codex$4.81 vs $10.00 per 1M blended; agentic coding session $0.161 vs $0.400–
Fastest first answerClaude Opus 4.835.4 s (Claude Opus 4.8 (max)) vs 55.6 s (GPT-5.3-Codex (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 4.8 (max) and GPT-5.3-Codex (xhigh) are neck and neck on quality (63.3 vs 63.8, within the ±6.9 margin).
  • Claude Opus 4.8 (max) is stronger in composite, knowledge, maths, reasoning.
  • GPT-5.3-Codex (xhigh) is stronger in agents & tools, coding, instruction following, long context, multimodal.
  • GPT-5.3-Codex (xhigh) is 2.1× cheaper ($4.81 vs $10.00 per 1M blended).
  • GPT-5.3-Codex (xhigh) streams 1.8× faster (105 vs 60 tokens per second).

At a glance

SpecClaude Opus 4.8GPT-5.3-Codex
DeveloperAnthropicOpenAI
Released28 May 20265 Feb 2026
Context window (tokens)1,000,000400,000
Compared atClaude Opus 4.8 (max)GPT-5.3-Codex (xhigh)
Open weightsNoNo
Price in / out / cached input per 1M$5.00 / $25.00 / $0.500$1.75 / $14.00 / $0.175

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 4.8 (max)GPT-5.3-Codex (xhigh)Leader
Reasoning68.868.6level
Coding58.960.9GPT-5.3-Codex
Agents & tools56.580.1GPT-5.3-Codex
Maths68.1–not comparable
Knowledge66.966.4level
Instruction following62.173.5GPT-5.3-Codex
Multimodal–61.5not comparable
Long context63.866.8GPT-5.3-Codex
Composite76.265.8Claude Opus 4.8

Where sources disagree

Across the 10 benchmarks both have, Claude Opus 4.8 leads 7, GPT-5.3-Codex leads 3. Every overall index that rates both puts Claude Opus 4.8 ahead.

Overall indexClaude Opus 4.8 (max)GPT-5.3-Codex (xhigh)Leader
BenchLeader Index63.363.8neck and neck
AA Intelligence Index v4.3.241.832.5Claude Opus 4.8

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 4.8 (max)GPT-5.3-Codex (xhigh)
Blended price per 1M tokens$10.00$4.81
Input / output per 1M$5.00 / $25.00$1.75 / $14.00
Cached input per 1M$0.500$0.175
Price per task: chat reply$0.0095$0.0049
Price per task: summarize a 30-page report$0.075$0.029
Price per task: code edit$0.068$0.032
Price per task: agentic coding session$0.400$0.161
Price per task: structured extraction$0.015$0.0063
Cost to run a full benchmark suite$4.08–
Output speed60 tok/s105 tok/s
Time to first answer35.4 s55.6 s
Time to a complete answer43.8 s60.4 s

Each model is compared at its best-ranked reasoning-effort setting. Claude Opus 4.8 (max) scores 63.3 and takes 35 s to its first answer token; Claude Opus 4.8 (high) scores 60.7 in 1 s, Claude Opus 4.8 scores 60.1 in 1 s. GPT-5.3-Codex (xhigh) scores 63.8 and takes 56 s to its first answer token. 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

124 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 4.8 (max)GPT-5.3-Codex (xhigh)
BenchLeader IndexBenchLeader63.363.8
Agents & tools scoreBenchLeader56.580.1
Coding scoreBenchLeader58.960.9
Composite scoreBenchLeader76.265.8
Instruction following scoreBenchLeader62.173.5
Knowledge scoreBenchLeader66.966.4
Long context scoreBenchLeader63.866.8
Maths scoreBenchLeader68.1–
Reasoning scoreBenchLeader68.868.6
Multimodal scoreBenchLeader–61.5
Blended price $/MBenchLeader$10.00$4.81
Output speedBenchLeader60 tok/s105 tok/s
Time to first answerBenchLeader35.4 s55.6 s
Context windowBenchLeader1,000,000 tokens400,000 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub91.0%–
SimpleBenchSimpleBench64.8%†–
Arena Hard PromptsArena (LMArena)1505†–
LiveBench 1.0 ReasoningLiveBench89.2%–
GPQA Diamond (AA)Artificial Analysis92.0%91.5%
Humanity's Last Exam (AA)Artificial Analysis48.7%42.5%
GPQA Diamond (Vals)Vals AI92.4%†–
EnigmaEvalScale AI SEAL23.5%†–
Kagi LLM BenchmarkKagi LLM Benchmark88.8%†–
ARC-AGI-1ARC Prize92.5%–
ARC-AGI-2ARC Prize72.1%†–
ARC-AGI-3ARC Prize1.5%†–
CritPtArtificial Analysis20.9%16.9%
Chess PuzzlesEpoch AI Benchmarking Hub52.0%–
Mystery Game PuzzlesEpoch AI Benchmarking Hub36.0%–
LMCAEpoch AI Benchmarking Hub57.5%–
DTBenchEpoch AI Benchmarking Hub94.9%–
ForecastBenchForecastBench66.3%–
BTF-3Epoch AI Benchmarking Hub14.0%†–
Coding
SWE-bench Verified (Epoch)Epoch AI Benchmarking Hub–74.8%†
SciCodeSciCode53.5%–
WeirdMLWeirdML70.5%†77.9%
FrontierCodeCognition46.5%†–
GSO-BenchGSO-Bench47.1%†–
Arena CodingArena (LMArena)1529†–
Show all 124 benchmarks
MetricPublisherClaude Opus 4.8 (max)GPT-5.3-Codex (xhigh)
Coding
Arena WebDevArena (LMArena)1536†1407†
LiveBench 1.0 CodingLiveBench81.8%–
SciCode (AA)Artificial Analysis54.4%–
LiveCodeBenchVals AI87.8%†87.3%
IOIVals AI–53.8%
SWE-bench (Vals)Vals AI88.6%†78.0%
Code MigrationVals AI47.3%†–
ProgramBenchVals AI1.0%†–
Vibe Code Bench v1.1Vals AI82.7%†61.8%
SWE Atlas: Codebase QnAScale AI SEAL57.3%†–
SWE Atlas: RefactoringScale AI SEAL46.7%†–
SWE Atlas: Test WritingScale AI SEAL49.6%†–
Surface Evolver BenchEpoch AI Benchmarking Hub87.5%†–
DeepSWE v1.1Epoch AI Benchmarking Hub59.0%–
ALE-BenchEpoch AI Benchmarking Hub1563.8†1655.2
Terminal-Bench 4.0 (AA)Artificial Analysis21.7%–
Terminal-Bench 2.1 (AA)Artificial Analysis84.6%–
Agents & tools
Terminal-BenchTerminal-Bench23.6%78.4%†
OSWorld-Verified 2.0OSWorld20.6%–
Remote Labor IndexScale AI / CAIS8.3%†–
APEX-AgentsMercor48.9%–
LiveBench Agentic CodingLiveBench50.5%–
Terminal-Bench HardArtificial Analysis58.3%53.0%
τ²-Bench Telecom (AA)Artificial Analysis94.4%86.0%
Terminal-Bench 2.1 (Vals)Vals AI71.9%†–
MCP AtlasScale AI SEAL82.2%–
HiL-BenchScale AI SEAL35.3%†4.3%†
τ³-bench Bankingτ²-bench39.7%–
τ³-Banking (AA)Artificial Analysis34.2%–
Analyst Agent (AA)Artificial Analysis45.0%–
Finance Agent v2Vals AI53.9%†–
Harvey's Legal Agent Benchmark v1Vals AI9.6%†–
Legal Research BenchVals AI43.8%†–
SkillsBenchVals AI59.2%†–
Tax Agent BenchVals AI30.0%†–
Terminal-Bench 2.0 (Vals)Vals AI70.0%†64.0%†
Terminal-Bench 4.0 (Vals)Vals AI23.2%†–
Terminal-Bench ScienceVals AI4.3%†–
Time Horizon Index: KSPVals AI11.8%†–
Arena SearchArena (LMArena)1195†–
DeepResearch BenchDeepResearch Bench50.2%†–
PostTrainBenchEpoch AI Benchmarking Hub32.9%–
METR Time HorizonsMETR–74.5%†
GBAEvalEpoch AI Benchmarking Hub70.9%†–
Vending-Bench 2Andon Labs2992.35940.1†
AutomationBenchArtificial Analysis45.6%–
GDP.pdf (AA)Artificial Analysis22.8%–
EnterpriseOps-GymArtificial Analysis44.0%–
AA-Briefcase v1.1Artificial Analysis1319–
FACTS SearchGoogle79.6%†–
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub80.0%–
FrontierMath Tier 4Epoch AI Benchmarking Hub56.1%–
OTIS Mock AIMEEpoch AI Benchmarking Hub98.3%–
ProofBenchVals AI69.0%–
LiveBench MathematicsLiveBench94.3%–
AIME 2026MathArena100.0%–
HMMT February 2026MathArena95.5%–
MathArena ApexMathArena81.3%–
Arena MathsArena (LMArena)1477†–
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub53.0%–
LiveBench 1.0 Data AnalysisLiveBench66.0%–
AA-OmniscienceArtificial Analysis28.810.9
MMLU-ProVals AI89.6%†–
LegalBenchVals AI83.6%†–
CorpFinVals AI66.7%†–
TaxEvalVals AI75.6%†–
Excel Modeling BenchmarkVals AI69.4%†–
MedCodeVals AI53.2%†–
MedScribeVals AI85.8%†–
Public Benefits Bench v1.1Vals AI68.1%†–
DrugDiscoveryBenchScale AI SEAL46.8%–
EBR-benchEpoch AI Benchmarking Hub28.6%–
GDP.pdfEpoch AI Benchmarking Hub24.0%–
AA-Omniscience: accuracyArtificial Analysis48.8%52.9%
AA-Omniscience: non-hallucinationArtificial Analysis60.8%10.8%
Instruction following
LiveBench LanguageLiveBench79.7%–
LiveBench Instruction FollowingLiveBench72.0%–
IFBenchArtificial Analysis62.2%75.4%
Arena Instruction FollowingArena (LMArena)1478†–
Human preference
Arena TextArena (LMArena)1474†–
EQ-Bench 4EQ-Bench1267†–
Arena Creative WritingArena (LMArena)1463†–
Arena Multi-turnArena (LMArena)1495†–
Arena Longer QueriesArena (LMArena)1499†–
Multimodal
Arena VisionArena (LMArena)1289†–
MMMU-ProArtificial Analysis–78.5%
MMMU-Pro (Vals)Vals AI86.6%†–
MortgageTaxVals AI69.9%†–
SAGEVals AI54.8%†–
Blueprint-Bench 2Epoch AI Benchmarking Hub14.5%†–
Long context
AA-LCRArtificial Analysis77.7%83.3%
Arena DocumentArena (LMArena)1475†–
MLCRArtificial Analysis19.4%–
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub159.0†157.1†
LiveBench 1.0LiveBench76.2%–
AA Intelligence Index v4.3.2Artificial Analysis41.832.5
Vals RSI IndexVals AI24.0%–
Vals Multimodal IndexVals AI70.9%†–
Safety & honesty
FORTRESSScale AI SEAL18.2%–

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 4.8 (max): 1 new result (AA-Briefcase v1.1).
  • Claude Opus 4.8 (max): 1 new result (Vals RSI Index).
  • Claude Opus 4.8 (max): 1 score revision.
  • Claude Opus 4.8 (max): 1 new result (ForecastBench).
  • Claude Opus 4.8 (max): BenchLeader Index 62.5 → 63.4.
  • Claude Opus 4.8 (max): 5 new results (AutomationBench, GDP.pdf (AA), MLCR…).
  • Claude Opus 4.8 (max): 1 score revision.
  • Claude Opus 4.8 (max): 1 new result (DrugDiscoveryBench).
  • GPT-5.3-Codex (xhigh): BenchLeader Index 64.4 → 63.9.
  • Claude Opus 4.8 (max): 1 score revision.

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

Claude Opus 4.8 vs GPT-5.3-Codex: questions

Is Claude Opus 4.8 better than GPT-5.3-Codex?
Claude Opus 4.8 (max) and GPT-5.3-Codex (xhigh) are neck and neck on quality (63.3 vs 63.8, within the ±6.9 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 2026-10-26, so check the category scores for your use.
Which is cheaper, Claude Opus 4.8 or GPT-5.3-Codex, and will that change?
GPT-5.3-Codex is cheaper today: $4.81 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 4.8 better than GPT-5.3-Codex for coding?
GPT-5.3-Codex scores higher in coding (61 vs 59 on the category index, where 50 is average).
Is Claude Opus 4.8 better than GPT-5.3-Codex for agents and tool use?
GPT-5.3-Codex scores higher in agentic tasks (80 vs 57 on the category index, where 50 is average). On Terminal-Bench Hard, the top shared agents & tools benchmark, it is 58.3% for Claude Opus 4.8 against 53.0% for GPT-5.3-Codex.
Which is better for long documents?
GPT-5.3-Codex scores higher on long-context benchmarks (67 vs 64). Claude Opus 4.8 accepts more context: 1,000,000 against 400,000 tokens for GPT-5.3-Codex.
Which has lower latency?
Claude Opus 4.8 (max) answers first: 35.4 s against 55.6 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 4.8 or GPT-5.3-Codex?
GPT-5.3-Codex streams faster: 105 against 60 output tokens per second.
Do independent benchmarks agree on Claude Opus 4.8 vs GPT-5.3-Codex?
Across the 10 benchmarks both have, Claude Opus 4.8 leads 7, GPT-5.3-Codex leads 3. Every overall index that rates both puts Claude Opus 4.8 ahead.
When was Claude Opus 4.8 released?
Claude Opus 4.8 was released on 28 May 2026.