Inference
EN

GPT-6 Astra vs Muse Wonder Full

Change models
GPT-6 AstraOpenAI
Muse Wonder FullMeta
IndexAll categories, out of 100.
78.0 ±1.9 best
77.3 ±5.3
GapIndex points below the top.
Top
−0.7 · no clear gap
CoverageIndex weight that has results.
100% best
80%
SpeedOutput tokens per second.
34/s
—
EloLMArena rating and rank.
1440
—
ReleasedMonth of release.
2026-09
2026-10
Input / 1MUS dollars per 1M input tokens.
$10.00
—
Output / 1MUS dollars per 1M output tokens.
$50.00
—
Batch output / 1MBatch price per 1M output tokens.
$25.00
—
1K chats2K tokens in, 500 out, each.
$45.00
—
1K documents20K tokens in, 1K out, each.
$250.00
—
1K agent tasks200K tokens in, 20K out, each.
$3,000.00
—
ContextMaximum tokens in one request.
1.05M best
1M
AgenticMulti-step tasks with tools.
74.6
78.2 best
CodingCode writing and repair.
76.9
77.3 best
ReasoningLogic problems and puzzles.
83.8 best
75.6
MultimodalTasks with images and text.
71.2
—
KnowledgeFacts and expert knowledge.
78.5 best
71.7
MultilingualTasks in many languages.
73.6
—
InstructionTasks with strict prompt rules.
79.5
—
MathMath problems.
78.0
79.6 best

Benchmarks

37 shared
Only GPT-6 Astra: 101Only Muse Wonder Full: 5
GPT-6 Astra
Muse Wonder Full
Artificial Analysis Humanity's Last ExamKnowledge
54.7% best8 / 222
53.6%9 / 222
Artificial Analysis Long Context ReasoningReasoning
80.7%37 / 217
83.7% best10 / 217
Critical Physics TasksReasoning
31.7%4 / 214
32.9% best2 / 214
Artificial Analysis Omniscience AccuracyKnowledge
61.6% best6 / 164
41.0%46 / 164
GDPval-AA normalizedAgentic
53.7%25 / 133
67.9% best2 / 133
Vibe Code Bench v1.1Coding
89.6% best8 / 123
86.0%14 / 123
Artificial Analysis SciCodeCoding
56.5%24 / 114
62.6% best4 / 114
MedScribeHealthcare
87.9%16 / 109
89.3% best8 / 109
MedCodeHealthcare
48.5%29 / 106
49.9% best18 / 106
Legal Research BenchLegal
39.4%37 / 90
53.4% best6 / 90
Tax Agent BenchFinance
20.7%52 / 82
43.0% best8 / 82
Finance Agent (v2)Finance
53.5%33 / 79
59.1% best6 / 79
Code MigrationCoding
67.7% best3 / 77
61.0%7 / 77
EMBFinance
71.7%11 / 75
71.9% best10 / 75
ProgramBenchCoding
5.5% best5 / 65
0.5%19 / 65
Terminal-Bench 4.0Agentic
59.6% best5 / 61
58.1%6 / 61
DeepSWEAgentic
74.1%11 / 59
79.0% best1 / 59
Agent Arena steerabilityAgentic
14.8 best5 / 54
9.612 / 54
Agent Arena task outcomeAgentic
14.0 best6 / 54
11.19 / 54
ProofBench v1.1Math
99.0% best6 / 54
94.0%11 / 54