BenchLeader
Menu
Best LLMsChoose a modelCoding plansLeaderboardsModelsBenchmarksProvidersReleases12

Claude Opus 5.5 vs Claude Sonnet 5

Full profiles: Claude Opus 5.5 (Anthropic) and Claude Sonnet 5 (Anthropic): every benchmark, price and rank.

Updated · Data from 12 sources

As of 9 Nov 2026, Claude Opus 5.5 is ahead of Claude Sonnet 5 on quality (68.9 vs 59.2 on the BenchLeader Index), but Claude Sonnet 5 is 2.0× cheaper at $4.00 vs $8.00 per 1M tokens blended. Claude Opus 5.5 leads in every category we measure both in. Claude Opus 5.5 streams faster.

Key facts (as of 9 Nov 2026)

  • Claude Opus 5.5 scores 68.9 on the BenchLeader Index and Claude Sonnet 5 scores 59.2 (BenchLeader, 9 Nov 2026).
  • Claude Opus 5.5 costs $8.00 per 1M tokens blended (list price, 9 Nov 2026).
  • Claude Sonnet 5 costs $4.00 per 1M tokens blended (list price, 9 Nov 2026).
  • Claude Opus 5.5 streams faster than Claude Sonnet 5 (Artificial Analysis, 7-day median, 9 Nov 2026).
  • Of the 14 benchmarks both models have, Claude Opus 5.5 leads 14 (BenchLeader, 9 Nov 2026).
  • On Arena WebDev, Claude Opus 5.5 is rated 1818 and Claude Sonnet 5 1539 (LMArena, as published by 9 Nov 2026).
Claude Opus 5.5 benchmarks max Anthropic Claude Sonnet 5 benchmarks high AnthropicAdd a third model

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 5.580 vs 60 category score; Arena WebDev 1818 vs 1539High
Agents & tool useClaude Opus 5.563 vs 54 category score; Arena Agent 20 vs 2.1High
MathsClaude Opus 5.574 vs 65 category scoreHigh
ReasoningClaude Opus 5.572 vs 66 category score; Humanity's Last Exam (AA) 61.4% vs 35.7%Medium
Long documentsClaude Opus 5.567 vs 63 category score; AA-LCR v1.1 84.7% vs 76.7%Medium
Images and other mediaClaude Opus 5.570 vs 61 category scoreHigh
Chat quality (human preference)–Only Claude Sonnet 5 has an Arena rating (65 category score)Not comparable
Lowest costClaude Sonnet 5$4.00 vs $8.00 per 1M blended; agentic coding session $0.160 vs $0.320–
Fastest first answerClaude Sonnet 515.6 s (Claude Sonnet 5 (high)) vs 380.6 s (Claude Opus 5.5 (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 5.5 (max) leads on quality: 68.9 vs 59.2.
  • Claude Opus 5.5 (max) is stronger in agents & tools, coding, knowledge, long context, maths, multimodal, reasoning.
  • Claude Sonnet 5 (high) is stronger in human preference.
  • Claude Sonnet 5 (high) is 2.0× cheaper ($4.00 vs $8.00 per 1M blended).

At a glance

SpecClaude Opus 5.5Claude Sonnet 5
DeveloperAnthropicAnthropic
Released22 Sep 202630 Jun 2026
Context window (tokens)1,000,0001,000,000
Max output (tokens)128,000 (up to 300K on the Message Batches API (beta))–
Input typestext, image–
Reasoning effort settingslow, medium, high, xhigh, max–
Compared atClaude Opus 5.5 (max)Claude Sonnet 5 (high)
Open weightsNoNo
Price in / out / cached input per 1M$4.00 / $20.00 / $0.200$2.00 / $10.00 / $0.200
API model IDclaude-opus-5-5–
Where to use itClaude API, Amazon Bedrock, Google Cloud, Microsoft Foundry–

Limits, availability and announced prices from the vendors' own pages: platform.claude.com (checked 20 Oct 2026). Everything else is from the day's data.

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 5.5 (max)Claude Sonnet 5 (high)Leader
Reasoning72.365.5Claude Opus 5.5
Coding80.460.4Claude Opus 5.5
Agents & tools62.754.3Claude Opus 5.5
Maths74.065.3Claude Opus 5.5
Knowledge73.7–not comparable
Human preference–64.7not comparable
Multimodal70.360.9Claude Opus 5.5
Long context67.162.9Claude Opus 5.5

On one chart

Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.

2550 avg75ReasoningCodingAgents & toolsMathsLong contextMultimodalClaude Opus 5.5, Reasoning: 72.3Claude Opus 5.5, Coding: 80.4Claude Opus 5.5, Agents & tools: 62.7Claude Opus 5.5, Maths: 74.0Claude Opus 5.5, Long context: 67.1Claude Opus 5.5, Multimodal: 70.3Claude Sonnet 5, Reasoning: 65.5Claude Sonnet 5, Coding: 60.4Claude Sonnet 5, Agents & tools: 54.3Claude Sonnet 5, Maths: 65.3Claude Sonnet 5, Long context: 62.9Claude Sonnet 5, Multimodal: 60.9

Not drawn, measured for fewer than two of these models: Knowledge, Instruction following, Human preference.

ModelIndex

Where sources disagree

Across the 14 benchmarks both have, Claude Opus 5.5 leads 14, Claude Sonnet 5 leads 0. Every overall index that rates both puts Claude Opus 5.5 ahead.

Overall indexClaude Opus 5.5 (max)Claude Sonnet 5 (high)Leader
BenchLeader Index68.959.2Claude Opus 5.5
Arena Text1506†1461Claude Opus 5.5

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. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.

MeasureClaude Opus 5.5 (max)Claude Sonnet 5 (high)
Blended price per 1M tokens$8.00$4.00
Input / output per 1M$4.00 / $20.00$2.00 / $10.00
Cached input per 1M$0.200$0.200
Price per task: chat reply$0.0076$0.0038
Price per task: summarize a 30-page report$0.060$0.030
Price per task: code edit$0.054$0.027
Price per task: agentic coding session$0.320$0.160
Price per task: structured extraction$0.012$0.0060
Agentic coding session, input mostly cached$0.149$0.079
Cost per benchmark task (AA Index)$10.04–
Output tokens per task (AA Index)173k–
Of which reasoning120k–
Time per task (AA Index)25 min–
Output speed115 tok/s105 tok/s
Time to first answer380.6 s15.6 s
Time to a complete answer771.1 s48.9 s

Each model is compared at its best-ranked reasoning-effort setting. Claude Opus 5.5 (max) scores 68.9 and takes 381 s to its first answer token; Claude Opus 5.5 (xhigh) scores 65.5 in 84 s, Claude Opus 5.5 (high) scores 65.4 in 28 s. Claude Sonnet 5 (high) scores 59.2 and takes 16 s to its first answer token; Claude Sonnet 5 scores 56.0 in 2 s, Claude Sonnet 5 (xhigh) scores 55.5 in 25 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

158 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 5.5 (max)Claude Sonnet 5 (high)
BenchLeader IndexBenchLeader68.959.2
Agents & tools scoreBenchLeader62.754.3
Coding scoreBenchLeader80.460.4
Knowledge scoreBenchLeader73.7–
Long context scoreBenchLeader67.162.9
Maths scoreBenchLeader74.065.3
Multimodal scoreBenchLeader70.360.9
Reasoning scoreBenchLeader72.365.5
Human preference scoreBenchLeader–64.7
Blended price $/MBenchLeader$8.00$4.00
Output speedBenchLeader115 tok/s105 tok/s
Time to first answerBenchLeader380.6 s15.6 s
Context windowBenchLeader1,000,000 tokens1,000,000 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub90.6%90.5%†
SimpleBenchSimpleBench–60.6%†
Arena Hard PromptsArena (LMArena)1530†1490
LiveBench 2.0 ReasoningLiveBench92.2%†–
LiveBench 1.0 ReasoningLiveBench92.2%88.7%†
GPQA Diamond (AA)Artificial Analysis–91.1%†
Humanity's Last Exam (AA)Artificial Analysis61.4%35.7%
HLE-Diamond (AA)Artificial Analysis67.9%–
GPQA Diamond (Vals)Vals AI–88.9%†
ARC-AGI-1ARC Prize97.5%–
ARC-AGI-2ARC Prize91.7%–
CritPtArtificial Analysis31.7%15.1%
CritPt v2 hard subset (AA)Artificial Analysis68.0%–
MysteryMechanismVals AI49.5%†–
Chess PuzzlesEpoch AI Benchmarking Hub67.0%34.0%†
Mystery Game PuzzlesEpoch AI Benchmarking Hub71.0%16.0%†
LMCAEpoch AI Benchmarking Hub68.2%50.0%
DTBenchEpoch AI Benchmarking Hub98.9%84.5%
ForecastBenchForecastBench–63.5%†
BTF-3Epoch AI Benchmarking Hub–14.2%†
Coding
SciCodeSciCode66.9%54.3%
WeirdMLWeirdML–68.8%
FrontierCodeCognition54.6%†42.7%†
GSO-BenchGSO-Bench–37.3%†
Arena CodingArena (LMArena)1546†1520
Show all 158 benchmarks
MetricPublisherClaude Opus 5.5 (max)Claude Sonnet 5 (high)
Coding
Arena WebDevArena (LMArena)18181539
LiveBench 2.0 CodingLiveBench87.1%†–
LiveBench 1.0 CodingLiveBench89.3%80.7%†
LiveCodeBenchVals AI–82.4%†
IOIVals AI95.1%†45.0%†
SWE-bench (Vals)Vals AI–79.6%†
Code MigrationVals AI66.7%†44.4%†
Code Migration v1.1Vals AI81.0%†–
ProgramBenchVals AI18.5%†0.0%†
SWE-bench MultimodalVals AI70.2%†–
Vibe Code Bench 1-100Vals AI30.4%†13.8%†
Vibe Code Bench v1.1Vals AI90.3%†81.3%†
MirrorCodeEpoch AI Benchmarking Hub78.9%†–
Surface Evolver BenchEpoch AI Benchmarking Hub–60.0%†
DeepSWE v1.1Epoch AI Benchmarking Hub–48.2%
CursorBenchCursor57.8%30.8%
ALE-BenchEpoch AI Benchmarking Hub2147.4†1463.1
FrontierSWEEpoch AI Benchmarking Hub62.3%–
Terminal-Bench 4.0 (AA)Artificial Analysis59.6%5.0%
Terminal-Bench 2.1 (AA)Artificial Analysis–80.5%†
Startrise FrontendStartrise71.1%†66.6%†
GameBench 2 OverallDuelLab GameBench 287.6†46.4†
GameBench 2 by reasoning settingDuelLab GameBench 294.8†54.4†
Design Arena WebsiteDesign Arena1325†1270†
FrontierCode 1.1 (AA)Artificial Analysis52.2%–
FrontierCode 2.0 (AA)Artificial Analysis46.4%17.8%†
Agents & tools
Terminal-BenchTerminal-Bench65.8%12.1%†
Remote Labor IndexScale AI / CAIS21.3%†–
APEX-AgentsMercor73.5%54.5%†
Arena AgentArena (LMArena)202.1
LiveBench Long AgenticLiveBench72.8%†–
LiveBench BrowserLiveBench88.4%†–
LiveBench Agentic CodingLiveBench71.7%59.4%†
Terminal-Bench 2.1 (Vals)Vals AI87.6%†74.5%†
MCP AtlasScale AI SEAL85.4%†–
HiL-BenchScale AI SEAL59.0%†–
GDPval-AA v2.1, normalised scoreArtificial Analysis68.3%–
GDPval-AA v2.1Artificial Analysis1866–
τ³-Banking (AA)Artificial Analysis–37.3%†
ITBench SRE (AA)Artificial Analysis38.2%–
Analyst Agent (AA)Artificial Analysis56.3%46.3%†
CUA-benchVals AI14.0%–
CyberBenchVals AI55.4%†61.9%†
Finance Agent v2Vals AI58.6%†53.9%†
Finance Agent v2.1Vals AI67.8%†–
Harvey's Legal Agent Benchmark v1.1Vals AI9.2%†–
Harvey's Legal Agent Benchmark v1Vals AI3.8%†5.0%†
Legal Research BenchVals AI50.5%†41.8%†
Robo-benchVals AI14.6%–
SkillsBenchVals AI–46.5%†
SREBenchVals AI33.6%†0.4%†
Tax Agent BenchVals AI44.5%†29.4%†
Terminal-Bench 4.0 (Vals)Vals AI65.2%†10.6%†
Terminal-Bench ScienceVals AI47.1%†5.7%†
Time Horizon Index: KSPVals AI91.3%†–
Arena SearchArena (LMArena)1214†1191†
GBAEvalEpoch AI Benchmarking Hub–65.3%†
Vending-Bench 2Andon Labs9235.3†6377.7†
AutomationBenchArtificial Analysis69.5%32.1%
GDP.pdf (AA)Artificial Analysis26.2%9.4%
Harvey LAB-AA v1.1Harvey4.2%–
EnterpriseOps-GymArtificial Analysis–43.9%†
AA-Briefcase v1.1Artificial Analysis18001356†
GDPxGDPx by SemiAnalysis12.0%–
Design Arena AgenticDesign Arena1334†1217†
Terminal-Bench-Science 0.1 (AA)Artificial Analysis59.0%–
CWE-Bench (AA)Artificial Analysis58.3%–
DeepSec-Bench (AA)Artificial Analysis26.9%–
CyberGym end-to-end (AA)Artificial Analysis0.8%–
FACTS SearchGoogle88.4%†–
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub91.2%65.6%†
FrontierMath Tier 4Epoch AI Benchmarking Hub95.0%29.3%†
OTIS Mock AIMEEpoch AI Benchmarking Hub100.0%94.7%†
ProofBenchVals AI100.0%77.0%†
LiveBench MathematicsLiveBench97.1%92.9%†
Arena MathsArena (LMArena)1518†1472
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub72.2%32.9%†
LiveBench 2.0 Data AnalysisLiveBench80.3%†–
LiveBench ResearchLiveBench88.9%†–
LiveBench 1.0 Data AnalysisLiveBench80.3%71.7%†
AA-Omniscience v1.1Artificial Analysis48.1–
MMLU-ProVals AI–87.5%†
LegalBenchVals AI–83.9%†
CorpFinVals AI–67.0%†
TaxEvalVals AI–75.6%†
PRBench LegalScale AI SEAL47.6%†–
BioMysteryBenchVals AI79.3%†–
ClaimBenchVals AI100.0%†–
Excel Modeling BenchmarkVals AI75.9%†66.3%†
Excel Modeling Benchmark v1.1Vals AI75.8%†–
MedCodeVals AI49.8%†47.5%†
MedScribeVals AI91.4%†76.0%†
Public Benefits Bench v1.1Vals AI70.6%†66.0%†
DrugDiscoveryBenchScale AI SEAL–50.0%†
EBR-benchEpoch AI Benchmarking Hub71.4%–
GDP.pdfEpoch AI Benchmarking Hub30.6%–
AA-Omniscience v1.1: accuracyArtificial Analysis65.5%–
AA-Omniscience v1.1: non-hallucinationArtificial Analysis49.7%–
HealthBench Professional (AA)OpenAI59.9%–
Instruction following
LiveBench LanguageLiveBench86.3%75.0%†
LiveBench Instruction FollowingLiveBench65.7%63.9%†
Arena Instruction FollowingArena (LMArena)1519†1467
Human preference
Arena TextArena (LMArena)1506†1461
EQ-Bench 4EQ-Bench1340†1221†
Arena Creative WritingArena (LMArena)1517†1434
Arena Multi-turnArena (LMArena)1500†1473
Arena Longer QueriesArena (LMArena)1520†1481
Multimodal
Arena VisionArena (LMArena)1324†1272
LiveBench DesignLiveBench83.3%†–
MMMU-ProArtificial Analysis87.7%77.3%†
MMMU-Pro (Vals)Vals AI–83.0%†
MortgageTaxVals AI–70.0%†
SAGEVals AI91.3%†–
Blueprint-Bench 2Epoch AI Benchmarking Hub51.2%†24.9%†
Code2Video-AAArtificial Analysis1536–
AA-3DArtificial Analysis1202–
Long context
AA-LCR v1.1Artificial Analysis84.7%76.7%
Arena DocumentArena (LMArena)–1469
MLCRArtificial Analysis27.8%4.4%
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub168.4†156.7†
LiveBench 2.0LiveBench83.4%†–
LiveBench 1.0LiveBench83.2%76.0%†
AA Intelligence Index v5.0Artificial Analysis59.1–
AA Intelligence Index v4.3.2Artificial Analysis57.631.7
AA Progress Index v1Artificial Analysis350296.8
Vals IndexVals AI70.1†–
Vals Finance IndexVals AI65.3%†–
Vals RSI IndexVals AI37.3%–
Vals Science IndexVals AI58.6%†–
Vals Multimodal IndexVals AI–68.8%†
Vals Router IndexVals AI64.2%†–
AA Cyber IndexArtificial Analysis28.7–
Safety & honesty
FORTRESSScale AI SEAL5.8%–

Data as of 2026-11-09. Best configuration of each model; every score links to its source on the model pages.

DuelLab GameBench 2, release gb2-20261103-eaea31df94e6 (https://benchmarks.duellab.org/releases/gb2-20261103-eaea31df94e6/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.

What changed

  • Claude Opus 5.5 (max): 2 score revisions.
  • Claude Opus 5.5 (max): 1 score revision.
  • Claude Sonnet 5 (high): 1 score revision.
  • Claude Opus 5.5 (max): 1 score revision.
  • Claude Opus 5.5 (max): 2 score revisions.
  • Claude Opus 5.5 (max): 2 score revisions.
  • Claude Opus 5.5 (max): 1 score revision.
  • Claude Sonnet 5 (high): 1 score revision.
  • Claude Opus 5.5 (max): 5 new results (AA Intelligence Index v5.0, AA Progress Index v1, HLE-Diamond (AA)…); 3 score revisions; cost per task moved from $5.98/M to $10.04/M (68% up).
  • Claude Sonnet 5 (high): 1 new result (AA Progress Index v1).

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

Claude Opus 5.5 vs Claude Sonnet 5: questions

Is Claude Opus 5.5 better than Claude Sonnet 5?
Claude Opus 5.5 (max) leads on quality: 68.9 vs 59.2. The BenchLeader Index combines every independent quality benchmark; Claude Opus 5.5 (max) is ahead overall as of 9 Nov 2026, but check the category scores for your use.
Claude Opus 5.5 vs Claude Sonnet 5: which is cheaper, and will that change?
Claude Sonnet 5 is cheaper today: $4.00 against $8.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Is Claude Opus 5.5 or Claude Sonnet 5 better for coding?
Claude Opus 5.5 scores higher in coding (80 vs 60 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1818 for Claude Opus 5.5 against 1539 for Claude Sonnet 5.
Which is faster, Claude Opus 5.5 or Claude Sonnet 5?
Claude Opus 5.5 streams faster (115 vs 105 output tokens per second, Artificial Analysis 7-day median).
Is Claude Opus 5.5 or Claude Sonnet 5 better for agents and tool use?
Claude Opus 5.5 scores higher in agentic tasks (63 vs 54 on the category index, where 50 is average). On Arena Agent, the top shared agents & tools benchmark, it is 20 for Claude Opus 5.5 against 2.1 for Claude Sonnet 5.
More questions (4)
Which is better for long documents?
Claude Opus 5.5 scores higher on long-context benchmarks (67 vs 63). Both accept 1,000,000 tokens of context.
Which has lower latency?
Claude Sonnet 5 answers first: 15.6 s against 380.6 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.
Do independent benchmarks agree on Claude Opus 5.5 vs Claude Sonnet 5?
Across the 14 benchmarks both have, Claude Opus 5.5 leads 14, Claude Sonnet 5 leads 0. Every overall index that rates both puts Claude Opus 5.5 ahead.
When was Claude Opus 5.5 released?
Claude Opus 5.5 was released on 22 Sep 2026.