Inference
EN

Gemini 4 Carbon vs Muse Spark 1.3

Change models
Gemini 4 CarbonGoogle
Muse Spark 1.3Meta
IndexAll categories, out of 100.
82.7 ±5.1 best
70.1 ±4.0
GapIndex points below the top.
Top
−12.6
CoverageIndex weight that has results.
80%
85% best
SpeedOutput tokens per second.
31/s
76/s best
EloLMArena rating and rank.
1538 best
1491
ReleasedMonth of release.
2026-10
2026-09
Input / 1MUS dollars per 1M input tokens.
$2.00
$1.25 best
Output / 1MUS dollars per 1M output tokens.
$10.00
$4.25 best
Batch output / 1MBatch price per 1M output tokens.
$5.00
—
1K chats2K tokens in, 500 out, each.
$9.00
$4.63 best
1K documents20K tokens in, 1K out, each.
$50.00
$29.25 best
1K agent tasks200K tokens in, 20K out, each.
$600.00
$335.00 best
ContextMaximum tokens in one request.
1.05M
1.05M
AgenticMulti-step tasks with tools.
83.2 best
70.4
CodingCode writing and repair.
80.5 best
68.8
ReasoningLogic problems and puzzles.
83.9 best
68.1
MultimodalTasks with images and text.
—
—
KnowledgeFacts and expert knowledge.
79.5 best
67.5
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
—
84.4
MathMath problems.
77.2 best
68.6

Benchmarks

17 shared
Only Gemini 4 Carbon: 22Only Muse Spark 1.3: 42
Gemini 4 Carbon
Muse Spark 1.3
Artificial Analysis Long Context ReasoningReasoning
79.3%52 / 217
83.0% best18 / 217
OTIS Mock AIME 2024-2025Math
100.0% best1 / 200
99.2%19 / 200
Artificial Analysis Omniscience AccuracyKnowledge
58.3% best9 / 164
43.6%41 / 164
Vibe Code Bench v1.1Coding
89.1% best9 / 123
82.9%23 / 123
Chess PuzzlesReasoning
78.0% best2 / 90
37.0%39 / 90
FrontierMath-Tiers-1-3-v2-PrivateMath
90.5% best6 / 90
74.0%25 / 90
Legal Research BenchLegal
55.8% best1 / 90
40.9%33 / 90
Tax Agent BenchFinance
51.8% best1 / 82
37.9%14 / 82
Mystery Game PuzzlesReasoning
73.0% best5 / 80
25.0%37 / 80
FrontierMath-Tier-4-v2-PrivateMath
95.1% best4 / 72
46.3%24 / 72
Terminal-Bench 4.0Agentic
63.1% best4 / 61
10.6%45 / 61
Agent Arena steerabilityAgentic
19.1 best2 / 54
2.622 / 54
Agent Arena task outcomeAgentic
18.4 best4 / 54
3.720 / 54
AutomationBenchAgentic
53.9% best5 / 32
49.4%14 / 32
FrontierSWE v2Coding
61.2% best5 / 22
25.8%15 / 22
cursorBench40Coding
60.4% best1 / 21
41.6%14 / 21
CWE-bench v1Agentic
71.0% best1 / 12
55.0%8 / 12