Inference
EN

Muse Spark vs Muse Spark 1.1

Change models
Muse SparkMeta
Muse Spark 1.1Meta
IndexAll categories, out of 100.
59.8 ±3.5
64.6 ±3.1 best
GapIndex points below the top.
−4.8
Top
CoverageIndex weight that has results.
95%
95%
SpeedOutput tokens per second.
—
152/s
EloLMArena rating and rank.
1474
1479 best
ReleasedMonth of release.
2026-04
2026-07
Input / 1MUS dollars per 1M input tokens.
—
$1.25
Output / 1MUS dollars per 1M output tokens.
—
$4.25
1K chats2K tokens in, 500 out, each.
—
$4.63
1K documents20K tokens in, 1K out, each.
—
$29.25
1K agent tasks200K tokens in, 20K out, each.
—
$335.00
ContextMaximum tokens in one request.
262K
1.05M best
AgenticMulti-step tasks with tools.
55.9
61.9 best
CodingCode writing and repair.
53.3
64.8 best
ReasoningLogic problems and puzzles.
59.8
63.0 best
MultimodalTasks with images and text.
63.0
66.5 best
KnowledgeFacts and expert knowledge.
67.8 best
67.4
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
60.0
67.5 best
MathMath problems.
60.2
61.2 best

Benchmarks

21 shared
Only Muse Spark: 26Only Muse Spark 1.1: 38
Muse Spark
Muse Spark 1.1
Artificial Analysis Humanity's Last ExamKnowledge
40.7%45 / 222
46.2% best24 / 222
Artificial Analysis Long Context ReasoningReasoning
78.0% best70 / 217
77.7%74 / 217
Critical Physics TasksReasoning
11.3%60 / 214
15.1% best50 / 214
Artificial Analysis GPQA DiamondKnowledge
88.4%51 / 199
89.8% best38 / 199
Artificial Analysis Omniscience AccuracyKnowledge
49.6%21 / 164
52.1% best14 / 164
LegalBenchLegal
84.2%37 / 149
85.0% best27 / 149
TaxEval v2Finance
77.7%3 / 143
79.7% best2 / 143
GPQA Diamond · Vals AIKnowledge
89.6%28 / 135
91.2% best23 / 135
MMLU ProKnowledge
87.3%29 / 135
88.7% best17 / 135
GDPval-AA normalizedAgentic
24.9%73 / 133
35.6% best50 / 133
CorpFin v2Finance
65.1%32 / 131
71.3% best4 / 131
Vibe Code Bench v1.1Coding
19.7%92 / 123
72.2% best46 / 123
MedScribeHealthcare
85.9%25 / 109
88.9% best9 / 109
MMMU ProMultimodal
87.4% best16 / 92
86.6%18 / 92
SWE-benchCoding
74.4%48 / 87
82.0% best21 / 87
SWE-bench ProCoding
52.4%64 / 83
61.5% best28 / 83
Humanity's Last ExamKnowledge
50.4%21 / 77
62.1% best6 / 77
Humanity's Last Exam without toolsKnowledge
42.8%24 / 53
52.2% best7 / 53
CharXiv ReasoningMultimodal
86.4%17 / 49
88.4% best13 / 49
CyberGymAgentic
43.5%28 / 29
59.0% best26 / 29