BenchLeader

GPT-5.4 vs GPT-6 Astra

Full profiles: GPT-5.4 (OpenAI) and GPT-6 Astra (OpenAI): every benchmark, price and rank.

Updated · Data from 15 sources

As of 10 Nov 2026, GPT-6 Astra is ahead of GPT-5.4 on quality (67.9 vs 63.2 on the BenchLeader Index), but GPT-5.4 is 3.6× cheaper at $5.63 vs $20.00 per 1M tokens blended. Choose GPT-5.4 for knowledge (59 vs 56), where it also streams ~2× faster; choose GPT-6 Astra for fast first answers (4.9 s vs 111.7 s) and reasoning (86 vs 66), coding (83 vs 65), and maths (77 vs 65).

Key facts (as of 10 Nov 2026)

  • GPT-5.4 scores 63.2 on the BenchLeader Index and GPT-6 Astra scores 67.9 (BenchLeader, 10 Nov 2026).
  • GPT-5.4 costs $5.63 per 1M tokens blended (list price, 10 Nov 2026).
  • GPT-6 Astra costs $20.00 per 1M tokens blended (list price, 10 Nov 2026).
  • GPT-5.4 streams ~2× faster than GPT-6 Astra (Artificial Analysis, 7-day median, 10 Nov 2026).
  • Of the 5 benchmarks both models have, GPT-6 Astra leads 5 (BenchLeader, 10 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-6 Astra83 vs 65 category score; GSO-Bench 79.4% vs 31.4%High
Agents & tool useGPT-6 Astra69 vs 63 category scoreMedium
MathsGPT-6 Astra77 vs 65 category score; ProofBench 99.0% vs 56.0%High
ReasoningGPT-6 Astra86 vs 66 category score; Humanity's Last Exam 54.8% vs 36.2%High
KnowledgeGPT-5.459 vs 56 category scoreMedium
Lowest costGPT-5.4$5.63 vs $20.00 per 1M blended; agentic coding session $0.210 vs $0.800–
Fastest first answerGPT-6 Astra4.9 s (GPT-6 Astra) vs 111.7 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

  • GPT-6 Astra leads on quality: 67.9 vs 63.2.
  • GPT-5.4 (xhigh) is stronger in instruction following, knowledge, long context, multimodal.
  • GPT-6 Astra is stronger in agents & tools, coding, maths, reasoning, composite.
  • GPT-5.4 (xhigh) is 3.6× cheaper ($5.63 vs $20.00 per 1M blended).
  • GPT-5.4 (xhigh) streams 2.8× faster (76 vs 27 tokens per second).

At a glance

SpecGPT-5.4GPT-6 Astra
DeveloperOpenAIOpenAI
Released5 Mar 20262 Sep 2026
Context window (tokens)1,050,0001,050,000
Compared atGPT-5.4 (xhigh)GPT-6 Astra
Open weightsNoNo
Price in / out / cached input per 1M$2.50 / $15.00 / $0.250$10.00 / $50.00 / $1.00

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-5.4 (xhigh)GPT-6 AstraLeader
Reasoning65.885.6GPT-6 Astra
Coding64.883.2GPT-6 Astra
Agents & tools63.068.8GPT-6 Astra
Maths65.177.2GPT-6 Astra
Knowledge58.656.1GPT-5.4
Instruction following72.4–not comparable
Multimodal60.9–not comparable
Long context65.7–not comparable
Composite–72.6not comparable

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 & toolsMathsKnowledgeGPT-5.4, Reasoning: 65.8GPT-5.4, Coding: 64.8GPT-5.4, Agents & tools: 63.0GPT-5.4, Maths: 65.1GPT-5.4, Knowledge: 58.6GPT-6 Astra, Reasoning: 85.6GPT-6 Astra, Coding: 83.2GPT-6 Astra, Agents & tools: 68.8GPT-6 Astra, Maths: 77.2GPT-6 Astra, Knowledge: 56.1

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

ModelIndex

Where sources disagree

Across the 5 benchmarks both have, GPT-5.4 leads 0, GPT-6 Astra leads 5. Independent sources split: GPT-5.4 leads on Arena Text; GPT-6 Astra leads on BenchLeader Index.

Overall indexGPT-5.4 (xhigh)GPT-6 AstraLeader
BenchLeader Index63.267.9GPT-6 Astra
Arena Text1475†1474GPT-5.4

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-5.4 (xhigh)GPT-6 Astra
Blended price per 1M tokens$5.63$20.00
Input / output per 1M$2.50 / $15.00$10.00 / $50.00
Cached input per 1M$0.250$1.00
Price per task: chat reply$0.0055$0.019
Price per task: summarize a 30-page report$0.039$0.150
Price per task: code edit$0.037$0.135
Price per task: agentic coding session$0.210$0.800
Price per task: structured extraction$0.0080$0.030
Agentic coding session, input mostly cached$0.109$0.395
Output speed76 tok/s27 tok/s
Time to first answer111.7 s4.9 s
Time to a complete answer195.8 s–

Each model is compared at its best-ranked reasoning-effort setting. GPT-5.4 (xhigh) scores 63.2 and takes 112 s to its first answer token; GPT-5.4 (low) scores 58.8 in 4 s, GPT-5.4 (high) scores 58.8 in 2 s. GPT-6 Astra scores 67.9 and takes 5 s to its first answer token; GPT-6 Astra (max) scores 67.3 in 239 s, GPT-6 Astra (high) scores 66.4 in 61 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

187 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-5.4 (xhigh)GPT-6 Astra
BenchLeader IndexBenchLeader63.267.9
Agents & tools scoreBenchLeader63.068.8
Coding scoreBenchLeader64.883.2
Instruction following scoreBenchLeader72.4–
Knowledge scoreBenchLeader58.656.1
Long context scoreBenchLeader65.7–
Maths scoreBenchLeader65.177.2
Multimodal scoreBenchLeader60.9–
Reasoning scoreBenchLeader65.885.6
Composite scoreBenchLeader–72.6
Blended price $/MBenchLeader$5.63$20.00
Output speedBenchLeader76 tok/s27 tok/s
Time to first answerBenchLeader111.7 s4.9 s
Context windowBenchLeader1,050,000 tokens1,050,000 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub93.3%95.8%†
Humanity's Last ExamScale AI / CAIS36.2%54.8%
Arena Hard PromptsArena (LMArena)1487†1500†
LiveBench 2.0 ReasoningLiveBench–92.7%
LiveBench 1.0 ReasoningLiveBench88.1%92.7%†
GPQA Diamond (AA)Artificial Analysis92.0%96.1%†
Humanity's Last Exam (AA)Artificial Analysis43.7%54.7%†
HLE-Diamond (AA)Artificial Analysis–63.8%†
GPQA Diamond (Vals)Vals AI91.7%–
Kagi LLM BenchmarkKagi LLM Benchmark63.8%†–
ARC-AGI-1ARC Prize93.7%97.5%†
ARC-AGI-2ARC Prize74.0%95.0%†
ARC-AGI-3ARC Prize0.2%†98.6%†
CritPtArtificial Analysis23.4%31.7%†
CritPt v2 hard subset (AA)Artificial Analysis–70.5%†
MysteryMechanismVals AI–53.1%†
Chess PuzzlesEpoch AI Benchmarking Hub48.0%74.0%†
Mystery Game PuzzlesEpoch AI Benchmarking Hub37.0%84.0%†
CL-benchEpoch AI Benchmarking Hub27.9%–
CL-bench LifeEpoch AI Benchmarking Hub21.7%–
LMCAEpoch AI Benchmarking Hub52.0%64.1%
DTBenchEpoch AI Benchmarking Hub94.4%97.1%
ForecastBenchForecastBench63.0%†66.6%†
BTF-3Epoch AI Benchmarking Hub–13.5%†
Coding
SWE-bench Verified (Epoch)Epoch AI Benchmarking Hub76.9%†–
Show all 187 benchmarks
MetricPublisherGPT-5.4 (xhigh)GPT-6 Astra
Coding
SciCodeSciCode56.6%56.5%†
WeirdMLWeirdML77.7%93.3%†
FrontierCodeCognition–53.3%†
GSO-BenchGSO-Bench31.4%79.4%
Arena CodingArena (LMArena)1511†1548†
Arena WebDevArena (LMArena)1400†1767†
LiveBench 2.0 CodingLiveBench–87.5%
LiveBench 1.0 CodingLiveBench77.5%80.4%†
SciCode (AA)Artificial Analysis–56.5%†
LiveCodeBenchVals AI84.1%–
IOIVals AI–100.0%†
SWE-bench (Vals)Vals AI78.2%–
SWE-Bench ProScale AI SEAL59.1%–
Code MigrationVals AI35.0%67.7%†
Code Migration v1.1Vals AI–80.6%†
ProgramBenchVals AI0.0%5.5%†
SWE-bench MultimodalVals AI–43.3%†
Vibe Code Bench 1-100Vals AI–27.6%†
Vibe Code Bench v1.1Vals AI67.4%89.6%†
SWE-Bench Pro (private)Scale AI SEAL59.1%–
SWE Atlas: Codebase QnAScale AI SEAL40.8%59.1%†
SWE Atlas: RefactoringScale AI SEAL44.3%59.0%†
SWE Atlas: Test WritingScale AI SEAL44.4%50.7%†
MirrorCodeEpoch AI Benchmarking Hub15.6%†46.7%†
DeepSWE v1.1Epoch AI Benchmarking Hub51.8%73.2%†
ALE-BenchEpoch AI Benchmarking Hub1607†2951.3†
AlgoTuneAlgoTune1.9†–
FrontierSWEEpoch AI Benchmarking Hub–65.5%†
Terminal-Bench 4.0 (AA)Artificial Analysis–59.1%†
Terminal-Bench 2.1 (AA)Artificial Analysis78.3%88.4%†
Startrise FrontendStartrise–79.9%
GameBench 2 OverallDuelLab GameBench 253.4†88.7
GameBench 2 by reasoning settingDuelLab GameBench 261.894.0†
Design Arena WebsiteDesign Arena1194†1332†
FrontierCode 1.1 (AA)Artificial Analysis–47.0%†
FrontierCode 2.0 (AA)Artificial Analysis–42.4%†
Agents & tools
Terminal-BenchTerminal-Bench81.8%†60.3%†
Remote Labor IndexScale AI / CAIS–20.8%
APEX-AgentsMercor52.4%†64.7%
Arena AgentArena (LMArena)–14.2†
LiveBench Long AgenticLiveBench–75.6%
LiveBench BrowserLiveBench–91.3%
LiveBench Agentic CodingLiveBench53.8%57.3%†
Terminal-Bench HardArtificial Analysis57.6%–
τ²-Bench Telecom (AA)Artificial Analysis87.1%–
Terminal-Bench 2.1 (Vals)Vals AI–87.3%†
MCP AtlasScale AI SEAL70.6%–
HiL-BenchScale AI SEAL9.7%†65.3%
τ³-bench Bankingτ²-bench39.4%–
GDPval-AA v2.1, normalised scoreArtificial Analysis–53.7%†
GDPval-AA v2.1Artificial Analysis–1586†
τ³-Banking (AA)Artificial Analysis39.6%41.4%†
ITBench SRE (AA)Artificial Analysis–48.6%†
Analyst Agent (AA)Artificial Analysis–51.3%†
APEX-Agents (AA)Artificial Analysis33.3%–
CUA-benchVals AI–19.2%†
CyberBenchVals AI–41.1%†
Finance Agent v2Vals AI–53.5%†
Finance Agent v2.1Vals AI–60.4%†
Harvey's Legal Agent Benchmark v1.1Vals AI–7.1%†
Harvey's Legal Agent Benchmark v1Vals AI0.0%5.4%†
Legal Research BenchVals AI–39.4%†
Robo-benchVals AI–23.3%†
SkillsBenchVals AI51.7%–
SREBenchVals AI–56.9%†
Tax Agent BenchVals AI–20.7%†
Terminal-Bench 2.0 (Vals)Vals AI58.4%†–
Terminal-Bench 4.0 (Vals)Vals AI–59.6%†
Terminal-Bench ScienceVals AI–62.9%†
Time Horizon Index: KSPVals AI–90.5%†
Arena SearchArena (LMArena)1196†1245†
BALROGBALROG–68.3%†
DeepResearch BenchDeepResearch Bench35.1%†–
PostTrainBenchEpoch AI Benchmarking Hub19.0%†–
METR Time HorizonsMETR74.3%–
GBAEvalEpoch AI Benchmarking Hub45.1%†–
Vending-Bench 2Andon Labs6144.2†15514.7
AutomationBenchArtificial Analysis–68.5%†
GDP.pdf (AA)Artificial Analysis–31.0%†
Harvey LAB-AA v1.1Harvey–8.6%†
AA-Briefcase v1.1Artificial Analysis–1569†
GDPxGDPx by SemiAnalysis–9.6%†
Design Arena AgenticDesign Arena1035†1391†
RoboDojo (LLM as robot policy)RoboDojo–29.0
Terminal-Bench-Science 0.1 (AA)Artificial Analysis–63.3%†
CWE-Bench (AA)Artificial Analysis–63.3%†
DeepSec-Bench (AA)Artificial Analysis–36.9%†
CyberGym end-to-end (AA)Artificial Analysis–0.0%†
DeepSearchQAGoogle63.7%†–
FACTS SearchGoogle–93.3%
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub78.6%93.7%†
FrontierMath Tier 4Epoch AI Benchmarking Hub49.0%97.6%†
OTIS Mock AIMEEpoch AI Benchmarking Hub95.3%100.0%†
ProofBenchVals AI56.0%99.0%
LiveBench MathematicsLiveBench94.2%96.8%†
AIME (Vals)Vals AI96.7%–
AIME 2026MathArena99.2%–
HMMT February 2026MathArena97.7%–
USAMO 2026MathArena95.2%–
MathArena ApexMathArena54.2%–
Arena MathsArena (LMArena)1459†1482†
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub45.1%75.6%†
LiveBench 2.0 Data AnalysisLiveBench–83.0%
LiveBench ResearchLiveBench–90.6%
LiveBench 1.0 Data AnalysisLiveBench79.3%83.0%†
AA-Omniscience v1.1Artificial Analysis–42.4†
MMLU-ProVals AI87.5%–
LegalBenchVals AI86.0%–
CorpFinVals AI65.3%–
TaxEvalVals AI74.0%–
MedQAVals AI96.1%–
PRBench FinanceScale AI SEAL45.6%†47.5%
PRBench LegalScale AI SEAL44.4%†48.4%
MultiNRCScale AI SEAL58.3%†–
BioMysteryBenchVals AI–79.3%†
CaseLaw v2Vals AI63.8%–
Excel Modeling BenchmarkVals AI–71.7%†
Excel Modeling Benchmark v1.1Vals AI–72.7%†
MedCodeVals AI41.3%48.5%†
MedScribeVals AI77.5%87.9%†
DrugDiscoveryBenchScale AI SEAL–68.7%†
EBR-benchEpoch AI Benchmarking Hub25.4%76.2%†
GDP.pdfEpoch AI Benchmarking Hub–34.2%†
AA-Omniscience v1.1: accuracyArtificial Analysis–61.6%†
AA-Omniscience v1.1: non-hallucinationArtificial Analysis–50.2%†
HealthBench Professional (AA)OpenAI–66.0%†
Instruction following
LiveBench LanguageLiveBench82.6%89.4%†
LiveBench Instruction FollowingLiveBench70.2%75.6%†
IFBenchArtificial Analysis74.0%–
Arena Instruction FollowingArena (LMArena)1460†1474†
Human preference
Arena TextArena (LMArena)1465†1474†
EQ-Bench 4EQ-Bench1258†–
Arena Creative WritingArena (LMArena)1433†1450†
Arena Multi-turnArena (LMArena)1480†1483†
Arena Longer QueriesArena (LMArena)1477†1486†
Multimodal
Arena VisionArena (LMArena)1292†1279†
LiveBench DesignLiveBench–80.0%
MMMU-ProArtificial Analysis78.4%86.9%†
VISTAScale AI SEAL50.9%†–
MMMU-Pro (Vals)Vals AI87.5%–
MortgageTaxVals AI68.3%–
VTBScale AI SEAL29.2%†–
Blueprint-Bench 2Epoch AI Benchmarking Hub27.1%†49.7%
Code2Video-AAArtificial Analysis–1427†
AA-3DArtificial Analysis–1168†
Long context
AA-LCR v1.1Artificial Analysis82.0%80.7%†
Arena DocumentArena (LMArena)1471†1474†
MLCRArtificial Analysis–41.7%†
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub157.8†169.2
LiveBench 2.0LiveBench–84.5%
LiveBench 1.0LiveBench78.0%82.2%†
AA Intelligence Index v5.0Artificial Analysis–56.9†
AA Intelligence Index v4.3.2Artificial Analysis39.052.7†
AA Progress Index v1Artificial Analysis311.6349.6†
Vals IndexVals AI–65.0†
Vals Finance IndexVals AI–55.6%†
Vals RSI IndexVals AI15.5%27.1%†
Vals Science IndexVals AI–65.1%†
Vals Router IndexVals AI–65.0%
AA Cyber IndexArtificial Analysis–33.4†
Safety & honesty
FORTRESSScale AI SEAL–12.1%
MASKScale AI SEAL89.7%†–

Data as of 2026-11-10. 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

  • GPT-6 Astra: 5 score revisions.
  • GPT-5.4 (xhigh): BenchLeader Index 63.7 → 63.2.
  • GPT-6 Astra: 1 score revision.
  • GPT-6 Astra: 1 new result (FORTRESS); 1 score revision.
  • GPT-5.4 (xhigh): 1 score revision.
  • GPT-6 Astra: 1 score revision.
  • GPT-5.4 (xhigh): 1 new result (AA Progress Index v1).
  • GPT-6 Astra: 3 score revisions.
  • GPT-5.4 (xhigh): 2 new results (Harvey's Legal Agent Benchmark v1, Vals RSI Index).
  • GPT-6 Astra: 1 new result (Vals Router Index).

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

GPT-5.4 vs GPT-6 Astra: questions

Is GPT-6 Astra better than GPT-5.4?
GPT-6 Astra leads on quality: 67.9 vs 63.2. The BenchLeader Index combines every independent quality benchmark; GPT-6 Astra is ahead overall as of 10 Nov 2026, but check the category scores for your use.
GPT-5.4 vs GPT-6 Astra: which is cheaper, and will that change?
GPT-5.4 is cheaper today: $5.63 against $20.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Is GPT-5.4 or GPT-6 Astra better for coding?
GPT-6 Astra scores higher in coding (83 vs 65 on the category index, where 50 is average). On GSO-Bench, the top shared coding benchmark, it is 31.4% for GPT-5.4 against 79.4% for GPT-6 Astra.
Which is faster, GPT-5.4 or GPT-6 Astra?
GPT-5.4 streams ~2× faster (76 vs 27 output tokens per second, Artificial Analysis 7-day median).
Is GPT-5.4 or GPT-6 Astra better for agents and tool use?
GPT-6 Astra scores higher in agentic tasks (69 vs 63 on the category index, where 50 is average).
More questions (4)
Which is better for long documents?
Both accept 1,050,000 tokens of context.
Which has lower latency?
GPT-6 Astra answers first: 4.9 s against 111.7 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.
Do independent benchmarks agree on GPT-5.4 vs GPT-6 Astra?
Across the 5 benchmarks both have, GPT-5.4 leads 0, GPT-6 Astra leads 5. Independent sources split: GPT-5.4 leads on Arena Text; GPT-6 Astra leads on BenchLeader Index.
When was GPT-6 Astra released?
GPT-6 Astra was released on 2 Sep 2026.