BenchLeader

GPT-5.5 vs GPT-5.6 Luna

Full profiles: GPT-5.5 (OpenAI) and GPT-5.6 Luna (OpenAI): every benchmark, price and rank.

Updated · Data from 13 sources

As of 16 Nov 2026, GPT-5.5 is ahead of GPT-5.6 Luna on quality (64.1 vs 58.2 on the BenchLeader Index), but GPT-5.6 Luna is 25× cheaper at $0.450 vs $11.25 per 1M tokens blended. GPT-5.5 leads in every category we measure both in. GPT-5.6 Luna streams faster.

Key facts (as of 16 Nov 2026)

  • GPT-5.5 scores 64.1 on the BenchLeader Index and GPT-5.6 Luna scores 58.2 (BenchLeader, 16 Nov 2026).
  • GPT-5.5 costs $11.25 per 1M tokens blended (list price, 16 Nov 2026).
  • GPT-5.6 Luna costs $0.450 per 1M tokens blended (list price, 16 Nov 2026).
  • GPT-5.6 Luna streams faster than GPT-5.5 (Artificial Analysis, 7-day median, 16 Nov 2026).
  • Of the 22 benchmarks both models have, GPT-5.5 leads 18 (BenchLeader, 16 Nov 2026).
  • On Arena WebDev, GPT-5.5 is rated 1512 and GPT-5.6 Luna 1519 (LMArena, as published by 16 Nov 2026).

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-5.561 vs 56 category score; SciCode 55.8% vs 50.5%Medium
Agents & tool useGPT-5.562 vs 57 category score; τ³-Banking (AA) 39.0% vs 28.7%Medium
MathsGPT-5.569 vs 66 category scoreMedium
ReasoningGPT-5.568 vs 52 category score; ARC-AGI-2 85.0% vs 47.6%High
Long documentsGPT-5.567 vs 66 category score; AA-LCR v1.1 84.3% vs 81.7%Medium
Images and other mediaGPT-5.562 vs 60 category score; MMMU-Pro 79.9% vs 78.5%Medium
Chat quality (human preference)–Only GPT-5.6 Luna has an Arena rating (64 category score)Not comparable
Lowest costGPT-5.6 Luna$0.450 vs $11.25 per 1M blended; agentic coding session $0.017 vs $0.420–
Fastest first answerGPT-5.6 Luna28.1 s (GPT-5.6 Luna (xhigh)) vs 38.3 s (GPT-5.5 (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

  • GPT-5.5 (xhigh) leads on quality: 64.1 vs 58.2.
  • GPT-5.5 (xhigh) is stronger in agents & tools, coding, long context, maths, multimodal, reasoning.
  • GPT-5.6 Luna (xhigh) is 25× cheaper ($0.450 vs $11.25 per 1M blended).
  • GPT-5.6 Luna (xhigh) streams 1.4× faster (110 vs 81 tokens per second).

At a glance

SpecGPT-5.5GPT-5.6 Luna
DeveloperOpenAIOpenAI
Released23 Apr 20269 Jul 2026
Context window (tokens)1,050,0001,050,000
Compared atGPT-5.5 (xhigh)GPT-5.6 Luna (xhigh)
Open weightsNoNo
Price in / out / cached input per 1M$5.00 / $30.00 / $0.500$0.200 / $1.20 / $0.020

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.

CategoryGPT-5.5 (xhigh)GPT-5.6 Luna (xhigh)Leader
Reasoning68.151.8GPT-5.5
Coding60.756.4GPT-5.5
Agents & tools61.857.1GPT-5.5
Maths69.365.7GPT-5.5
Knowledge63.6–not comparable
Instruction following74.0–not comparable
Human preference–63.7not comparable
Multimodal62.359.9GPT-5.5
Long context66.965.5GPT-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 contextMultimodalGPT-5.5, Reasoning: 68.1GPT-5.5, Coding: 60.7GPT-5.5, Agents & tools: 61.8GPT-5.5, Maths: 69.3GPT-5.5, Long context: 66.9GPT-5.5, Multimodal: 62.3GPT-5.6 Luna, Reasoning: 51.8GPT-5.6 Luna, Coding: 56.4GPT-5.6 Luna, Agents & tools: 57.1GPT-5.6 Luna, Maths: 65.7GPT-5.6 Luna, Long context: 65.5GPT-5.6 Luna, Multimodal: 59.9

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

Model, click to add or removeIndex

Where sources disagree

Across the 22 benchmarks both have, GPT-5.5 leads 18, GPT-5.6 Luna leads 4. Every overall index that rates both puts GPT-5.5 ahead.

Overall indexGPT-5.5 (xhigh)GPT-5.6 Luna (xhigh)Leader
BenchLeader Index64.158.2GPT-5.5
Arena Text1481†1453GPT-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.

MeasureGPT-5.5 (xhigh)GPT-5.6 Luna (xhigh)
Blended price per 1M tokens$11.25$0.450
Input / output per 1M$5.00 / $30.00$0.200 / $1.20
Cached input per 1M$0.500$0.020
Price per task: chat reply$0.011$0.0004
Price per task: summarize a 30-page report$0.078$0.0031
Price per task: code edit$0.075$0.0030
Price per task: agentic coding session$0.420$0.017
Price per task: structured extraction$0.016$0.0006
Agentic coding session, input mostly cached$0.217$0.0087
Output speed81 tok/s110 tok/s
Time to first answer38.3 s28.1 s
Time to a complete answer, last measured 13 Nov 2026104.3 s101.7 s

Each model is compared at its best-ranked reasoning-effort setting. GPT-5.5 (xhigh) scores 64.1 and takes 38 s to its first answer token; GPT-5.5 (high) scores 63.5 in 16 s, GPT-5.5 scores 61.0 in 3 s. GPT-5.6 Luna (xhigh) scores 58.2 and takes 28 s to its first answer token; GPT-5.6 Luna (max) scores 56.2 in 59 s, GPT-5.6 Luna (high) scores 54.7 in 11 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

160 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.

MetricPublisherGPT-5.5 (xhigh)GPT-5.6 Luna (xhigh)
BenchLeader IndexBenchLeader64.158.2
Agents & tools scoreBenchLeader61.857.1
Coding scoreBenchLeader60.756.4
Instruction following scoreBenchLeader74.0–
Knowledge scoreBenchLeader63.6–
Long context scoreBenchLeader66.965.5
Maths scoreBenchLeader69.365.7
Multimodal scoreBenchLeader62.359.9
Reasoning scoreBenchLeader68.151.8
Human preference scoreBenchLeader–63.7
Blended price $/MBenchLeader$11.25$0.450
Output speedBenchLeader81 tok/s110 tok/s
Time to first answerBenchLeader38.3 s28.1 s
Context windowBenchLeader1,050,000 tokens1,050,000 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub90.7%†91.6%†
SimpleBenchSimpleBench69.0%†46.8%
Arena Hard PromptsArena (LMArena)1499†1475
LiveBench 1.0 ReasoningLiveBench89.7%85.6%†
GPQA Diamond (AA)Artificial Analysis93.5%89.5%
Humanity's Last Exam (AA)Artificial Analysis45.8%37.0%
HLE-Diamond (AA)Artificial Analysis–20.9%†
GPQA Diamond (Vals)Vals AI93.2%91.7%†
Kagi LLM BenchmarkKagi LLM Benchmark88.8%†49.1%†
ARC-AGI-1ARC Prize95.0%87.7%
ARC-AGI-2ARC Prize85.0%47.6%
ARC-AGI-3ARC Prize0.4%†0.0%
CritPtArtificial Analysis27.1%20.6%
CritPt v2 hard subset (AA)Artificial Analysis–30.5%†
MysteryMechanismVals AI–14.4%†
Arena ExpertArena (LMArena)1510†1490
Chess PuzzlesEpoch AI Benchmarking Hub63.0%43.0%†
Mystery Game PuzzlesEpoch AI Benchmarking Hub56.0%21.0%†
CL-bench LifeEpoch AI Benchmarking Hub22.2%†–
LMCAEpoch AI Benchmarking Hub54.3%48.5%
DTBenchEpoch AI Benchmarking Hub96.0%89.1%
ForecastBenchForecastBench64.2%†–
BTF-3Epoch AI Benchmarking Hub14.3%†–
Coding
SciCodeSciCode55.8%50.5%
WeirdMLWeirdML84.9%60.9%†
Show all 160 benchmarks
MetricPublisherGPT-5.5 (xhigh)GPT-5.6 Luna (xhigh)
Coding
FrontierCodeCognition43.0%†39.8%†
GSO-BenchGSO-Bench40.2%–
Arena CodingArena (LMArena)1510†1503
Arena WebDevArena (LMArena)15121519
LiveBench 1.0 CodingLiveBench82.2%82.9%†
LiveCodeBenchVals AI85.3%–
IOIVals AI–61.8%†
SWE-bench (Vals)Vals AI82.6%93.0%†
Code MigrationVals AI45.2%44.5%†
ProgramBenchVals AI0.5%0.0%†
Vibe Code Bench 1-100Vals AI–22.6%†
Vibe Code Bench v1.1Vals AI69.8%77.1%†
SWE Atlas v1.0: Codebase QnAScale AI SEAL45.4%–
SWE Atlas v1.0: RefactoringScale AI SEAL44.8%–
SWE Atlas v1.0: Test WritingScale AI SEAL42.6%–
MirrorCodeEpoch AI Benchmarking Hub23.3%21.1%†
Surface Evolver BenchEpoch AI Benchmarking Hub88.1%†61.9%†
DeepSWE v1.1Epoch AI Benchmarking Hub67.0%56.9%
CursorBenchCursor–33.0%
ALE-BenchEpoch AI Benchmarking Hub1943.01667.4†
WeirdML v3WeirdML–5.2%
Terminal-Bench 4.0 (AA)Artificial Analysis14.7%3.5%
Terminal-Bench 2.1 (AA)Artificial Analysis84.3%77.9%
Startrise FrontendStartrise–65.0%†
GameBench 2 OverallDuelLab GameBench 249.2†40.2†
GameBench 2 by reasoning settingDuelLab GameBench 259.954
Design Arena WebsiteDesign Arena1248†1248†
FrontierCode 2.0 (AA)Artificial Analysis–17.6%†
Agents & tools
Terminal-BenchTerminal-Bench84.7%†17.6%†
OSWorld-Verified 2.0OSWorld13.0%–
Remote Labor IndexScale AI / CAIS6.3%†–
APEX-AgentsMercor55.1%†43.0%†
LiveBench Agentic CodingLiveBench54.0%48.4%†
Terminal-Bench HardArtificial Analysis60.6%–
τ²-Bench Telecom (AA)Artificial Analysis93.9%–
Terminal-Bench 2.1 (Vals)Vals AI57.3%†79.0%†
MCP AtlasScale AI SEAL75.3%–
HiL-BenchScale AI SEAL39.7%†–
τ³-bench Bankingτ²-bench44.6%–
GDPval-AA v2.1Artificial Analysis13571407
τ³-Banking (AA)Artificial Analysis39.0%28.7%
ITBench SRE (AA)Artificial Analysis45.8%40.3%†
Analyst Agent (AA)Artificial Analysis50.0%–
APEX-Agents (AA)Artificial Analysis37.7%35.8%†
CyberBenchVals AI–73.6%†
Finance Agent v2Vals AI51.8%55.0%†
Harvey's Legal Agent Benchmark v1Vals AI3.8%1.3%†
Legal Research BenchVals AI40.4%36.5%†
SkillsBenchVals AI62.2%60.5%†
SREBenchVals AI3.8%1.5%†
Tax Agent BenchVals AI21.8%24.8%†
Terminal-Bench 2.0 (Vals)Vals AI73.2%†–
Terminal-Bench 4.0 (Vals)Vals AI–11.6%†
Terminal-Bench ScienceVals AI–0.0%†
Time Horizon Index: KSPVals AI8.3%–
Arena SearchArena (LMArena)1226†–
BALROGBALROG–45.6%†
DeepResearch BenchDeepResearch Bench54.0%†–
PostTrainBenchEpoch AI Benchmarking Hub23.9%–
ExploitBenchEpoch AI Benchmarking Hub47.4%†–
GBAEvalEpoch AI Benchmarking Hub53.2%†–
Vending-Bench 2Andon Labs7523.8†4094.7†
AutomationBenchArtificial Analysis47.3%42.6%
GDP.pdf (AA)Artificial Analysis21.2%23.8%
EnterpriseOps-GymArtificial Analysis46.6%40.8%†
AA-Briefcase v1.1Artificial Analysis11351267
Design Arena AgenticDesign Arena1113†1149†
RoboDojo (LLM as robot policy)RoboDojo1.1†–
Terminal-Bench-Science 0.1 (AA)Artificial Analysis–3.3%†
FACTS SearchGoogle–87.0%†
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub85.3%82.1%†
FrontierMath Tier 4Epoch AI Benchmarking Hub72.5%61.0%†
OTIS Mock AIMEEpoch AI Benchmarking Hub84.4%†98.3%†
ProofBenchVals AI50.0%60.0%†
LiveBench MathematicsLiveBench95.9%87.2%†
AIME 2026MathArena100.0%–
HMMT February 2026MathArena98.5%–
USAMO 2026MathArena98.2%–
MathArena ApexMathArena80.2%–
Arena MathsArena (LMArena)1500†1476
FrontierMath ErdősEpoch AI Benchmarking Hub0.0%–
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub63.0%41.0%†
LiveBench 1.0 Data AnalysisLiveBench81.6%78.0%†
AA-Omniscience v1.1Artificial Analysis–-17.1†
MMLU-ProVals AI88.1%86.0%†
LegalBenchVals AI86.5%84.0%†
CorpFinVals AI68.4%64.2%†
TaxEvalVals AI75.0%76.2%†
BioMysteryBenchVals AI–61.5%†
CaseLaw v2Vals AI66.2%–
Excel Modeling BenchmarkVals AI64.5%67.1%†
MedCodeVals AI49.1%42.4%†
MedScribeVals AI86.9%84.4%†
Public Benefits Bench v1.1Vals AI60.9%61.2%†
Access to Justice BenchmarkVals AI–65.6%†
Financial Coaching BenchVals AI–69.3%†
DrugDiscoveryBenchScale AI SEAL51.6%–
EBR-benchEpoch AI Benchmarking Hub34.3%–
GDP.pdfEpoch AI Benchmarking Hub26.0%22.7%†
AA-Omniscience v1.1: accuracyArtificial Analysis–39.1%†
AA-Omniscience v1.1: non-hallucinationArtificial Analysis–7.7%†
FACTS ParametricGoogle–50.6%†
Instruction following
LiveBench LanguageLiveBench87.4%72.6%†
LiveBench Instruction FollowingLiveBench70.7%60.1%†
IFBenchArtificial Analysis75.8%–
Arena Instruction FollowingArena (LMArena)1475†1443
Human preference
Arena TextArena (LMArena)1476†1453
EQ-Bench 4EQ-Bench1295†1146†
EQ-Bench Creative Writing v3EQ-Bench1845†1832†
Arena Creative WritingArena (LMArena)1448†1411
Arena Multi-turnArena (LMArena)1484†1458
Arena Longer QueriesArena (LMArena)1486†1452
Multimodal
Arena VisionArena (LMArena)1296†1256
MMMU-ProArtificial Analysis79.9%78.5%
MMMU-Pro (Vals)Vals AI88.3%85.0%†
MortgageTaxVals AI68.8%67.3%†
Humanity's Sixth SenseScale AI SEAL–21.6%†
Blueprint-Bench 2Epoch AI Benchmarking Hub36.2%†22.6%†
Furniture AssemblyEpoch AI Benchmarking Hub44.2%42.5%†
Code2Video-AAArtificial Analysis1000–
FACTS MultimodalGoogle–37.4%†
Long context
AA-LCR v1.1Artificial Analysis84.3%81.7%
Arena DocumentArena (LMArena)1479†1460
MLCRArtificial Analysis28.9%15.0%
FACTS GroundingGoogle–52.5%†
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub160.1†156.7†
LiveBench 1.0LiveBench80.2%73.6%†
AA Intelligence Index v5.0.1Artificial Analysis–23.8†
AA Intelligence Index v4.3.2Artificial Analysis38.434.6
AA Progress Index v1Artificial Analysis320305.8
Vals RSI IndexVals AI16.5%–
Vals Science IndexVals AI–25.3%†
Vals Multimodal IndexVals AI68.1%69.1%†
Safety & honesty
FORTRESSScale AI SEAL16.3%–
DistressBenchScale AI SEAL–72.7%†

Data as of 2026-11-16. Best configuration of each model; every score links to its source on the model pages.

DuelLab GameBench 2, release gb2-20261111-7b598bb13de4 (https://benchmarks.duellab.org/releases/gb2-20261111-7b598bb13de4/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.

What changed

  • GPT-5.5 (xhigh): 1 new result (GDPval-AA v2.1).
  • GPT-5.6 Luna (xhigh): 2 new results (GDPval-AA v2.1, AA-Briefcase v1.1).
  • GPT-5.5 (xhigh): BenchLeader Index 64.6 → 64.1.
  • GPT-5.5 (xhigh): 1 score revision.
  • GPT-5.5 (xhigh): 1 new result (MirrorCode).
  • GPT-5.6 Luna (xhigh): 1 score revision.
  • GPT-5.5 (xhigh): 1 score revision.
  • GPT-5.6 Luna (xhigh): 1 score revision.
  • GPT-5.5 (xhigh): 1 new result (AA Progress Index v1).
  • GPT-5.6 Luna (xhigh): 1 new result (AA Progress Index v1).

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

GPT-5.5 vs GPT-5.6 Luna: questions

Is GPT-5.5 better than GPT-5.6 Luna?
GPT-5.5 (xhigh) leads on quality: 64.1 vs 58.2. The BenchLeader Index combines every independent quality benchmark; GPT-5.5 (xhigh) is ahead overall as of 16 Nov 2026, but check the category scores for your use.
GPT-5.5 vs GPT-5.6 Luna: which is cheaper, and will that change?
GPT-5.6 Luna is cheaper today: $0.450 against $11.25 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Is GPT-5.5 or GPT-5.6 Luna better for coding?
GPT-5.5 scores higher in coding (61 vs 56 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1512 for GPT-5.5 against 1519 for GPT-5.6 Luna.
Which is faster, GPT-5.5 or GPT-5.6 Luna?
GPT-5.6 Luna streams faster (110 vs 81 output tokens per second, Artificial Analysis 7-day median).
Is GPT-5.5 or GPT-5.6 Luna better for agents and tool use?
GPT-5.5 scores higher in agentic tasks (62 vs 57 on the category index, where 50 is average). On τ³-Banking (AA), the top shared agents & tools benchmark, it is 39.0% for GPT-5.5 against 28.7% for GPT-5.6 Luna.
More questions (4)
Which is better for long documents?
GPT-5.5 scores higher on long-context benchmarks (67 vs 66). Both accept 1,050,000 tokens of context.
Which has lower latency?
GPT-5.6 Luna answers first: 28.1 s against 38.3 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 GPT-5.5 vs GPT-5.6 Luna?
Across the 22 benchmarks both have, GPT-5.5 leads 18, GPT-5.6 Luna leads 4. Every overall index that rates both puts GPT-5.5 ahead.
When was GPT-5.6 Luna released?
GPT-5.6 Luna was released on 9 Jul 2026.