BenchLeader

GPT-6.2 Astra vs Muse Wonder

Full profiles: GPT-6.2 Astra (OpenAI) and Muse Wonder (Meta): every benchmark, price and rank.

Updated · Data from 13 sources

As of 19 Nov 2026, GPT-6.2 Astra is ahead of Muse Wonder on quality (68.9 vs 64.2 on the BenchLeader Index), but Muse Wonder is 2.0× cheaper at $8.00 vs $16.00 per 1M tokens blended. Choose GPT-6.2 Astra for maths (73 vs 70); choose Muse Wonder for fast first answers (2.2 s vs 202.7 s) and agents & tools (76 vs 66) and knowledge (76 vs 74), where it also streams ~2× faster.

Key facts (as of 19 Nov 2026)

  • GPT-6.2 Astra scores 68.9 on the BenchLeader Index and Muse Wonder scores 64.2 (BenchLeader, 19 Nov 2026).
  • GPT-6.2 Astra costs $16.00 per 1M tokens blended (list price, 19 Nov 2026).
  • Muse Wonder costs $8.00 per 1M tokens blended (list price, 19 Nov 2026).
  • Muse Wonder streams ~2× faster than GPT-6.2 Astra (Artificial Analysis, 7-day median, 19 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
Agents & tool useMuse Wonder76 vs 66 category scoreHigh
MathsGPT-6.2 Astra73 vs 70 category scoreMedium
KnowledgeMuse Wonder76 vs 74 category scoreMedium
Chat quality (human preference)–Only Muse Wonder has an Arena rating (70 category score)Not comparable
Lowest costMuse Wonder$8.00 vs $16.00 per 1M blended; agentic coding session $0.320 vs $0.640–
Fastest first answerMuse Wonder2.2 s (Muse Wonder) vs 202.7 s (GPT-6.2 Astra (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

  • GPT-6.2 Astra (max) leads on quality: 68.9 vs 64.2.
  • GPT-6.2 Astra (max) is stronger in maths.
  • Muse Wonder is stronger in agents & tools, knowledge.
  • Muse Wonder is 2.0× cheaper ($8.00 vs $16.00 per 1M blended).
  • Muse Wonder streams 2.3× faster (112 vs 49 tokens per second).

At a glance

SpecGPT-6.2 AstraMuse Wonder
DeveloperOpenAIMeta
Released17 Nov 202612 Oct 2026
Context window (tokens)1,050,0001,048,576
Compared atGPT-6.2 Astra (max)Muse Wonder
Open weightsNoNo
Price in / out / cached input per 1M$8.00 / $40.00 / $0.400$4.00 / $20.00 / $0.200

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-6.2 Astra (max)Muse WonderLeader
Reasoning73.1–not comparable
Coding76.1–not comparable
Agents & tools66.475.7Muse Wonder
Maths72.570.1GPT-6.2 Astra
Knowledge74.275.5Muse Wonder
Human preference–69.6not comparable
Multimodal71.7–not comparable
Composite69.5–not 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 avg75Agents & toolsMathsKnowledgeGPT-6.2 Astra, Agents & tools: 66.4GPT-6.2 Astra, Maths: 72.5GPT-6.2 Astra, Knowledge: 74.2Muse Wonder, Agents & tools: 75.7Muse Wonder, Maths: 70.1Muse Wonder, Knowledge: 75.5

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

Model, click to add or removeIndex

Where sources disagree

Across the 0 benchmarks both have, GPT-6.2 Astra leads 0, Muse Wonder leads 0. Every overall index that rates both puts GPT-6.2 Astra ahead.

Overall indexGPT-6.2 Astra (max)Muse WonderLeader
BenchLeader Index68.964.2GPT-6.2 Astra
AA Intelligence Index v5.0.162.140.5†GPT-6.2 Astra
Vals Index67.764.9†GPT-6.2 Astra

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-6.2 Astra (max)Muse Wonder
Blended price per 1M tokens$16.00$8.00
Input / output per 1M$8.00 / $40.00$4.00 / $20.00
Cached input per 1M$0.400$0.200
Price per task: chat reply$0.015$0.0076
Price per task: summarize a 30-page report$0.120$0.060
Price per task: code edit$0.108$0.054
Price per task: agentic coding session$0.640$0.320
Price per task: structured extraction$0.024$0.012
Agentic coding session, input mostly cached$0.298$0.149
Cost per benchmark task (AA Index)$4.00–
Output tokens per task (AA Index)41k–
Of which reasoning26k–
Time per task (AA Index)14 min–
Output speed49 tok/s112 tok/s
Time to first answer202.7 s2.2 s

Each model is compared at its best-ranked reasoning-effort setting. GPT-6.2 Astra (max) scores 68.9 and takes 203 s to its first answer token; GPT-6.2 Astra (xhigh) scores 67.1 in 67 s, GPT-6.2 Astra (medium) scores 65.6 in 4 s. Muse Wonder scores 64.2 and takes 2 s to its first answer token; Muse Wonder (max) scores 60.1 in 21 s, Muse Wonder (high) scores 55.6 in 10 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

144 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-6.2 Astra (max)Muse Wonder
BenchLeader IndexBenchLeader68.964.2
Agents & tools scoreBenchLeader66.475.7
Coding scoreBenchLeader76.1–
Composite scoreBenchLeader69.5–
Knowledge scoreBenchLeader74.275.5
Maths scoreBenchLeader72.570.1
Multimodal scoreBenchLeader71.7–
Reasoning scoreBenchLeader73.1–
Human preference scoreBenchLeader–69.6
Blended price $/MBenchLeader$16.00$8.00
Output speedBenchLeader49 tok/s112 tok/s
Time to first answerBenchLeader202.7 s2.2 s
Context windowBenchLeader1,050,000 tokens1,048,576 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub96.8%–
Arena Hard PromptsArena (LMArena)–1530†
LiveBench 2.0 ReasoningLiveBench91.6%90.6%†
LiveBench 1.0 ReasoningLiveBench–90.6%†
Humanity's Last Exam (AA)Artificial Analysis–50.1%†
HLE-Diamond (AA)Artificial Analysis72.2%39.0%†
ARC-AGI-1ARC Prize98.5%–
ARC-AGI-2ARC Prize95.8%–
ARC-AGI-3ARC Prize99.8%–
CritPtArtificial Analysis–28.3%†
CritPt v2 hard subset (AA)Artificial Analysis71.5%50.5%†
MysteryMechanismVals AI53.1%27.5%†
HLE-Diamond (Scale)Scale AI SEAL66.9%†37.3%
Arena ExpertArena (LMArena)–1525†
Chess PuzzlesEpoch AI Benchmarking Hub78.0%–
Mystery Game PuzzlesEpoch AI Benchmarking Hub95.0%51.0%†
LMCAEpoch AI Benchmarking Hub–56.8%
DTBenchEpoch AI Benchmarking Hub–95.2%
Coding
SciCodeSciCode–60.3%†
Arena CodingArena (LMArena)–1549†
Arena WebDevArena (LMArena)17811648†
LiveBench 2.0 CodingLiveBench91.9%74.9%†
LiveBench 1.0 CodingLiveBench–80.7%†
IOIVals AI–66.9%†
Code MigrationVals AI–62.5%†
Show all 144 benchmarks
MetricPublisherGPT-6.2 Astra (max)Muse Wonder
Coding
Code Migration v1.1Vals AI–72.4%†
SWE-bench MultimodalVals AI–45.6%†
Vibe Code Bench 1-100Vals AI–26.6%†
Vibe Code Bench v1.1Vals AI89.9%83.1%†
Vibe Code Bench 1-100 v1.1Vals AI55.2%–
SWE Atlas v1.1Scale AI SEAL60.5%†64.8%
SWE Atlas v1.1: Codebase QnAScale AI SEAL67.7%†73.4%
SWE Atlas v1.1: Test WritingScale AI SEAL62.2%†61.1%
SWE Atlas v1.1: RefactoringScale AI SEAL51.4%†60.0%
SWE-Bench Pro V2Scale AI SEAL–96.7%†
SWE-Bench Pro V2 HARDScale AI SEAL–90.2%†
MirrorCodeEpoch AI Benchmarking Hub67.8%45.6%†
CursorBenchCursor–50.4%†
WeirdML v3WeirdML–15.1%†
Terminal-Bench 4.0 (AA)Artificial Analysis64.1%52.0%†
GameBench 2 OverallDuelLab GameBench 2–64.8
GameBench 2 by reasoning settingDuelLab GameBench 2–54.9†
Design Arena WebsiteDesign Arena1339†1369†
FrontierCode 1.1 (AA)Artificial Analysis–46.5%†
FrontierCode 2.0 (AA)Artificial Analysis45.4%27.9%†
Agents & tools
Terminal-BenchTerminal-Bench62.1%47.0%†
APEX-AgentsMercor–66.1%†
Arena AgentArena (LMArena)15.20.6†
LiveBench Long AgenticLiveBench70.8%46.2%†
LiveBench BrowserLiveBench93.0%60.0%†
LiveBench Agentic CodingLiveBench–62.7%†
HiL-BenchScale AI SEAL–79.7%
GDPval-AA v2.1, normalised scoreArtificial Analysis–63.5%†
GDPval-AA v2.1Artificial Analysis–1770†
Analyst Agent (AA)Artificial Analysis62.5%–
CUA-benchVals AI–7.2%†
CyberBenchVals AI–80.4%†
Finance Agent v2Vals AI–56.6%†
Finance Agent v2.1Vals AI60.8%65.5%†
Harvey's Legal Agent Benchmark v1.1Vals AI16.3%23.8%†
Harvey's Legal Agent Benchmark v1Vals AI–17.1%†
Legal Research BenchVals AI42.8%50.5%†
Robo-benchVals AI32.7%–
Tax Agent BenchVals AI21.8%36.3%†
Terminal-Bench 4.0 (Vals)Vals AI63.1%49.5%†
Terminal-Bench ScienceVals AI71.4%22.9%†
Arena Agent Task CompletionArena (LMArena)12.42.5†
PostTrainBenchEpoch AI Benchmarking Hub–24.2%†
Vending-Bench 2Andon Labs–9432.2
SpireClimbEpoch AI Benchmarking Hub41.3%–
AutomationBenchArtificial Analysis77.3%68.1%†
GDP.pdf (AA)Artificial Analysis29.8%30.6%†
Harvey LAB-AA v1.1Harvey15.0%10.0%†
EnterpriseOps-GymArtificial Analysis29.9%–
AA-Briefcase v1.1Artificial Analysis16431659†
GDPxGDPx by SemiAnalysis–6.8%†
Design Arena AgenticDesign Arena13041173†
Terminal-Bench-Science 0.1 (AA)Artificial Analysis78.1%25.7%†
CWE-Bench (AA)Artificial Analysis65.0%–
DeepSec-Bench (AA)Artificial Analysis46.8%–
CyberGym end-to-end (AA)Artificial Analysis0.0%–
MKA (AA)Artificial Analysis52.7%47.5%†
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub92.6%82.1%†
FrontierMath Tier 4Epoch AI Benchmarking Hub100.0%70.7%†
OTIS Mock AIMEEpoch AI Benchmarking Hub100.0%99.4%†
ProofBenchVals AI–84.0%
LiveBench MathematicsLiveBench–96.5%†
Autoform-HardVals AI87.5%–
Arena MathsArena (LMArena)–1501†
FrontierMath ErdősEpoch AI Benchmarking Hub10.3%–
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub78.5%–
LiveBench 2.0 Data AnalysisLiveBench84.1%81.6%†
LiveBench ResearchLiveBench91.5%93.0%†
LiveBench 1.0 Data AnalysisLiveBench–81.6%†
AA-Omniscience v1.1Artificial Analysis5025.1†
PRBench FinanceScale AI SEAL–58.7%
PRBench LegalScale AI SEAL43.1%†58.4%
BioMysteryBenchVals AI80.7%67.0%†
Excel Modeling BenchmarkVals AI–67.1%†
Excel Modeling Benchmark v1.1Vals AI76.0%72.5%†
Public Benefits Bench v1.1Vals AI–70.2%†
Access to Justice BenchmarkVals AI–73.0%†
Financial Coaching BenchVals AI–73.8%†
Immigration BenchVals AI–69.6%†
DrugDiscoveryBenchScale AI SEAL65.5%†–
EBR-benchEpoch AI Benchmarking Hub92.4%31.4%†
GDP.pdfEpoch AI Benchmarking Hub–54.8%†
AA-Omniscience v1.1: accuracyArtificial Analysis68.6%38.0%†
AA-Omniscience v1.1: non-hallucinationArtificial Analysis40.9%79.1%†
Instruction following
LiveBench LanguageLiveBench–82.9%†
LiveBench Instruction FollowingLiveBench–73.0%†
MultiChallengeScale AI SEAL73.2%†–
Arena Instruction FollowingArena (LMArena)–1495†
Human preference
Arena TextArena (LMArena)–1511†
EQ-Bench 4EQ-Bench–1325
EQ-Bench Creative Writing v3EQ-Bench–2007
Arena Creative WritingArena (LMArena)–1488†
Arena Multi-turnArena (LMArena)–1522†
Arena Longer QueriesArena (LMArena)–1508†
Multimodal
LiveBench DesignLiveBench82.8%70.9%†
MMMU-ProArtificial Analysis89.5%83.8%†
Humanity's Sixth SenseScale AI SEAL63.2%†43.7%†
Blueprint-Bench 2Epoch AI Benchmarking Hub–31.7%
Furniture AssemblyEpoch AI Benchmarking Hub89.2%54.2%†
VIALS-AAHandshake AI65.8%55.3%†
Code2Video-AAArtificial Analysis–997†
AA-3DArtificial Analysis1318–
Long context
AA-LCR v1.1Artificial Analysis–80.7%†
MLCRArtificial Analysis39.4%22.2%†
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub174.2†161.7
LiveBench 2.0LiveBench84.9%70.7%†
LiveBench 1.0LiveBench–81.2%†
AA Intelligence Index v5.0.1Artificial Analysis62.143.6†
AA Intelligence Index v4.3.2Artificial Analysis–52.4†
AA Progress Index v1Artificial Analysis–335.8†
Vals IndexVals AI67.764.9†
Vals Finance IndexVals AI57.3%61.2%†
Vals RSI IndexVals AI29.0%–
Vals Science IndexVals AI68.4%39.1%†
Vals Router IndexVals AI–64.9%
Vals Legal IndexVals AI69.7%75.3%†
AA Cyber IndexArtificial Analysis37.3–
Safety & honesty
FORTRESSScale AI SEAL–13.8%
DistressBenchScale AI SEAL–85.9%

Data as of 2026-11-19. Best configuration of each model; every score links to its source on the model pages.

DuelLab GameBench 2, release gb2-20261114-ca113f475f5e (https://benchmarks.duellab.org/releases/gb2-20261114-ca113f475f5e/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.

What changed

  • GPT-6.2 Astra (max): BenchLeader Index 67.8 → 68.9.
  • GPT-6.2 Astra (max): 6 new results (VIALS-AA, Arena WebDev, Arena Agent…).
  • Muse Wonder: 1 new result (FORTRESS).
  • GPT-6.2 Astra (max): BenchLeader Index 67.2 → 67.8.
  • GPT-6.2 Astra (max): 2 new results (MKA (AA), Terminal-Bench).
  • Muse Wonder: BenchLeader Index 63.0 → 64.4.
  • Muse Wonder: BenchLeader Index 63.5 → 63.0.
  • Muse Wonder: 1 new result (Blueprint-Bench 2).
  • Muse Wonder: BenchLeader Index 64.3 → 63.5.
  • Muse Wonder: 5 score revisions.

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

GPT-6.2 Astra vs Muse Wonder: questions

Is GPT-6.2 Astra better than Muse Wonder?
GPT-6.2 Astra (max) leads on quality: 68.9 vs 64.2. The BenchLeader Index combines every independent quality benchmark; GPT-6.2 Astra (max) is ahead overall as of 19 Nov 2026, but check the category scores for your use.
GPT-6.2 Astra vs Muse Wonder: which is cheaper, and will that change?
Muse Wonder is cheaper today: $8.00 against $16.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Which is faster, GPT-6.2 Astra or Muse Wonder?
Muse Wonder streams ~2× faster (112 vs 49 output tokens per second, Artificial Analysis 7-day median).
Is GPT-6.2 Astra or Muse Wonder better for agents and tool use?
Muse Wonder scores higher in agentic tasks (76 vs 66 on the category index, where 50 is average).
Which is better for long documents?
Both take about 1.1M tokens: GPT-6.2 Astra 1,050,000, Muse Wonder 1,048,576.
More questions (2)
Which has lower latency?
Muse Wonder answers first: 2.2 s against 202.7 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.
When was GPT-6.2 Astra released?
GPT-6.2 Astra was released on 17 Nov 2026.