Inference
EN

Muse Spark 1.1 vs Qwen3.8-Flash-Next

Change models
Muse Spark 1.1Meta
Qwen3.8-Flash-NextAlibaba
IndexAll categories, out of 100.
64.6 ±3.1 best
64.3 ±3.2
GapIndex points below the top.
Top
−0.3 · no clear gap
CoverageIndex weight that has results.
95%
95%
SpeedOutput tokens per second.
152/s best
55/s
EloLMArena rating and rank.
1479
—
ReleasedMonth of release.
2026-07
2026-08
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.
1.05M best
262K
AgenticMulti-step tasks with tools.
61.9
68.5 best
CodingCode writing and repair.
64.8 best
62.1
ReasoningLogic problems and puzzles.
63.0 best
62.2
MultimodalTasks with images and text.
66.5 best
65.0
KnowledgeFacts and expert knowledge.
67.4 best
55.2
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
67.5
73.8 best
MathMath problems.
61.2 best
59.4

Benchmarks

24 shared
Only Muse Spark 1.1: 35Only Qwen3.8-Flash-Next: 25
Muse Spark 1.1
Qwen3.8-Flash-Next
Artificial Analysis Humanity's Last ExamKnowledge
46.2% best24 / 222
38.0%55 / 222
Artificial Analysis Long Context ReasoningReasoning
77.7%74 / 217
79.7% best45 / 217
Critical Physics TasksReasoning
15.1% best50 / 214
11.1%61 / 214
Artificial Analysis GPQA DiamondKnowledge
89.8%38 / 199
92.3% best19 / 199
Artificial Analysis Omniscience AccuracyKnowledge
52.1% best14 / 164
23.5%112 / 164
GDPval-AA normalizedAgentic
35.6%50 / 133
56.4% best13 / 133
Artificial Analysis SciCodeCoding
58.8% best14 / 114
50.6%49 / 114
SWE-bench ProCoding
61.5%28 / 83
62.5% best24 / 83
LiveBench Data AnalysisReasoning
72.5%48 / 80
74.2% best43 / 80
LiveBench ReasoningReasoning
87.7% best27 / 80
87.4%28 / 80
Humanity's Last ExamKnowledge
62.1% best6 / 77
35.9%39 / 77
LiveBench Agentic CodingAgentic
58.5%18 / 66
61.6% best12 / 66
LiveBench CodingCoding
77.2% best37 / 66
72.6%53 / 66
LiveBench Instruction FollowingInstruction
69.6%30 / 66
77.1% best5 / 66
LiveBench LanguageKnowledge
74.3%50 / 66
74.6% best49 / 66
LiveBench MathematicsMath
87.1% best44 / 66
85.8%50 / 66
DeepSWEAgentic
53.3%49 / 59
58.7% best44 / 59
Agent Arena steerabilityAgentic
−15.449 / 54
−2.6 best36 / 54
Agent Arena task outcomeAgentic
−6.540 / 54
0.3 best31 / 54
Humanity's Last Exam without toolsKnowledge
52.2% best7 / 53
35.9%32 / 53