Inference
EN

MiMo-V2.6-Flash vs Muse Spark 1.1

Change models
MiMo-V2.6-FlashXiaomi
Muse Spark 1.1Meta
IndexAll categories, out of 100.
64.3 ±4.4
64.6 ±3.1 best
GapIndex points below the top.
−0.3 · no clear gap
Top
CoverageIndex weight that has results.
90%
95% best
SpeedOutput tokens per second.
28/s
152/s best
EloLMArena rating and rank.
1450
1479 best
ReleasedMonth of release.
2026-09
2026-07
Input / 1MUS dollars per 1M input tokens.
$0.14 best
$1.25
Output / 1MUS dollars per 1M output tokens.
$0.28 best
$4.25
1K chats2K tokens in, 500 out, each.
$0.42 best
$4.63
1K documents20K tokens in, 1K out, each.
$3.08 best
$29.25
1K agent tasks200K tokens in, 20K out, each.
$33.60 best
$335.00
ContextMaximum tokens in one request.
1.05M
1.05M
AgenticMulti-step tasks with tools.
69.7 best
61.9
CodingCode writing and repair.
64.1
64.8 best
ReasoningLogic problems and puzzles.
59.2
63.0 best
MultimodalTasks with images and text.
50.3
66.5 best
KnowledgeFacts and expert knowledge.
52.3
67.4 best
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
—
67.5
MathMath problems.
68.9 best
61.2

Benchmarks

24 shared
Only MiMo-V2.6-Flash: 17Only Muse Spark 1.1: 35
MiMo-V2.6-Flash
Muse Spark 1.1
Artificial Analysis Humanity's Last ExamKnowledge
35.1%62 / 222
46.2% best24 / 222
Artificial Analysis Long Context ReasoningReasoning
74.3%98 / 217
77.7% best74 / 217
Critical Physics TasksReasoning
12.0%57 / 214
15.1% best50 / 214
Artificial Analysis Omniscience AccuracyKnowledge
27.2%91 / 164
52.1% best14 / 164
GDPval-AA normalizedAgentic
55.4% best19 / 133
35.6%50 / 133
Vibe Code Bench v1.1Coding
79.0% best33 / 123
72.2%46 / 123
Artificial Analysis SciCodeCoding
51.3%46 / 114
58.8% best14 / 114
MedScribeHealthcare
85.3%29 / 109
88.9% best9 / 109
Legal Research BenchLegal
38.0%43 / 90
38.0%43 / 90
Terminal-Bench 2.1 · BenchLMAgentic
87.6% best13 / 86
80.0%37 / 86
Finance Agent (v2)Finance
56.3%19 / 79
57.2% best17 / 79
Code MigrationCoding
40.9% best29 / 77
31.1%39 / 77
EMBFinance
65.5% best24 / 75
56.4%45 / 75
Terminal-Bench 2.1 · Vals AIAgentic
76.4% best16 / 75
69.3%30 / 75
ProgramBenchCoding
0.5% best19 / 65
0.0%34 / 65
DeepSWEAgentic
67.9% best26 / 59
53.3%49 / 59
Agent Arena steerabilityAgentic
2.8 best20 / 54
−15.449 / 54
Agent Arena task outcomeAgentic
4.2 best19 / 54
−6.540 / 54
Agent Arena command recoveryAgentic
−1.1 best36 / 50
−2.338 / 50
Public Benefits Bench v1.1Public benefits
67.6% best14 / 50
61.3%30 / 50