MetaReasoning modelCompare with another model

Muse Spark 1.3

Reasoning effort

Muse Spark 1.3 is a Meta proprietary reasoning model, released 2 Sept 2026. Its best configuration (max reasoning effort) ranks #26 of 775 on the BenchLeader Index at 67.2 ±4.4, in the upper half. The ± is the point: 86 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in composite (84) and lowest in maths (63). At $2.00 per million tokens blended it is pricier than most ranked models. Output speed of 128 tokens per second puts it faster than most, with a first answer in 80.4 s. It has been measured at 7 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 19 Oct 2026.

Blended price: $2.00/M; $1.25 in · $4.25 out
Output speed: 128 tok/s; measured by Artificial Analysis
First answer: 80 s; first token 3.04 s
Context: 1.0M
Full answer: 84 s; median, reasoning included
Cost per run: $1.60; one full Intelligence Index run
Released: 2 Sept 2026

How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.

Overall index67
Reasoning78
Coding65
Agents & tools66
Maths63
Knowledge73
Human preference69
Multimodal66
Long context67
Composite84

Versions

Meta has shipped 4 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.

Model	Released	Index	Rank
Muse Spark 1.3maxthis page	2 Sept 2026	67.2	#26
Muse Spark 1.2xhigh	5 Aug 2026	63.2	#69
Muse Spark 1.1	9 Jul 2026	62.9	#75
Muse Spark	8 Apr 2026	64.4	#50

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
minimal	–	–	87 tok/s	3.04 s	$0.0018
low	–	–	87 tok/s	3.04 s	$0.0018
medium	–	–	87 tok/s	3.04 s	$0.0018
high	–	–	87 tok/s	3.04 s	$0.0018
xhigh	64.9	#43	282 tok/s	26 s	$0.0018	Agents & tools 54 · Coding 60 · Composite 80 · Knowledge 72 · Long context 67 · Maths 62 · Multimodal 65 · Reasoning 84
maxbest	67.2	#26	128 tok/s	80 s	$0.0018	Agents & tools 66 · Coding 65 · Composite 84 · Human preference 69 · Knowledge 73 · Long context 67 · Maths 63 · Multimodal 66 · Reasoning 78
not stated	58.3	#158	87 tok/s	3.04 s	$0.0018	Agents & tools 59 · Composite 58 · Knowledge 80 · Maths 57

How its index has moved

9 Sept 2026 to 21 Oct 2026

max
xhigh
not stated

The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.

Benchmark results

See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Reasoning

Benchmark	xhigh	max	not stated	Source
Arena Hard Prompts	–	1515#12	–	Arena (LMArena)
LiveBench 2.0 Reasoningnot in indexeffort not stated	–	–	89.7%#14	LiveBench
LiveBench 1.0 Reasoningnot in index	89.7%#15	–	–	LiveBench
GPQA Diamond (AA)not in index	94.1%#7	93.5%#14	–	Artificial Analysis
Humanity's Last Exam (AA)not in index	47.5%#38	48.7%#31	–	Artificial Analysis
CritPt	26.0%#39	24.9%#44	–	Artificial Analysis
MysteryMechanismnot in index	–	36.0%#9	–	Vals AI
Chess Puzzlesnot in index	34.0%#57	37.0%#43	–	Epoch AI Benchmarking Hub
Mystery Game Puzzlesnot in index	14.1%#95	25.0%#46	–	Epoch AI Benchmarking Hub
LMCAnot in indexeffort not stated	–	–	53.9%#49	Epoch AI Benchmarking Hub
DTBenchnot in indexeffort not stated	–	–	96.5%#22	Epoch AI Benchmarking Hub
BTF-3not in index	14.1%#5	–	–	Epoch AI Benchmarking Hub

Coding

Benchmark	minimal	low	medium	high	xhigh	max	not stated	Source
SciCode	–	–	–	–	59.7%#12	58.8%#17	–	SciCode
Arena Coding	–	–	–	–	–	1535#15	–	Arena (LMArena)
Arena WebDev	–	–	–	–	1621#22	1659#13	–	Arena (LMArena)
LiveBench 2.0 Codingnot in indexeffort not stated	–	–	–	–	–	–	90.2%#12	LiveBench
LiveBench 1.0 Codingnot in index	–	–	–	–	81.1%#16	–	–	LiveBench
SciCode (AA)not in index	–	–	–	–	59.7%#13	58.8%#18	–	Artificial Analysis
IOI	–	–	–	–	43.9%#32	56.6%#19	–	Vals AI
Code Migrationnot in index	–	–	–	–	27.6%#44	47.4%#14	–	Vals AI
ProgramBenchnot in index	–	–	–	–	0.5%#18	2.5%#7	–	Vals AI
Vibe Code Bench 1-100not in index	–	–	–	–	–	20.5%#9	–	Vals AI
Vibe Code Bench v1.1not in index	–	–	–	–	82.9%#20	85.9%#14	–	Vals AI
CursorBenchnot in index	24.3%#75	29.3%#68	32.6%#60	33.4%#55	37.5%#44	41.6%#34	–	Cursor
ALE-Benchnot in index	–	–	–	–	1688.3#14	–	–	Epoch AI Benchmarking Hub
FrontierSWEnot in index	–	–	–	–	–	25.8%#13	–	Epoch AI Benchmarking Hub
Terminal-Bench 4.0 (AA)not in index	–	–	–	–	16.7%#67	33.3%#38	–	Artificial Analysis
Terminal-Bench 2.1 (AA)not in index	–	–	–	–	85.4%#25	84.3%#30	–	Artificial Analysis
Startrise Frontendnot in indexeffort not stated	–	–	–	–	–	–	74.7%#10	Startrise
FrontierCode 1.1 (AA)not in index	–	–	–	–	–	40.6%#24	–	Artificial Analysis

Agents & tools

Benchmark	xhigh	max	not stated	Source
Terminal-Bench	14.6%#78	–	–	Terminal-Bench
APEX-Agents	–	47.6%#31	58.6%#14	Mercor
Arena Agent	–	3.3#15	–	Arena (LMArena)
LiveBench Long Agenticnot in indexeffort not stated	–	–	40.4%#23	LiveBench
LiveBench Browsernot in indexeffort not stated	–	–	80.0%#16	LiveBench
LiveBench Agentic Codingnot in index	64.1%#8	–	–	LiveBench
Terminal-Bench 2.1 (Vals)	72.3%#24	79.0%#11	–	Vals AI
GDPval-AA v2.1not in index	56.4%#23	59.0%#16	–	Artificial Analysis
τ³-Banking (AA)	47.2%#8	50.5%#3	–	Artificial Analysis
ITBench SRE (AA)not in index	–	33.2%#29	–	Artificial Analysis
Analyst Agent (AA)not in index	–	37.5%#26	–	Artificial Analysis
CUA-benchnot in index	–	5.8%#7	–	Vals AI
CyberBenchnot in index	69.4%#26	72.7%#16	–	Vals AI
Finance Agent v2not in index	58.9%#8	60.0%#5	–	Vals AI
Legal Research Benchnot in index	40.9%#27	55.3%#1	–	Vals AI
SREBenchnot in index	–	3.8%#19	–	Vals AI
Tax Agent Benchnot in index	36.0%#12	42.4%#7	–	Vals AI
Terminal-Bench 4.0 (Vals)not in index	10.6%#37	24.8%#23	–	Vals AI
Terminal-Bench Sciencenot in index	4.3%#25	10.0%#15	–	Vals AI
AutomationBenchnot in index	56.8%#49	57.9%#46	–	Artificial Analysis
GDP.pdf (AA)not in index	24.2%#35	26.6%#20	–	Artificial Analysis
Harvey LAB-AA v1.1not in index	–	8.9%#4	–	Harvey
AA-Briefcase v1.1not in index	–	1581#16	–	Artificial Analysis
Terminal-Bench-Science 0.1 (AA)not in index	–	10.9%#24	–	Artificial Analysis

Maths

Benchmark	xhigh	max	not stated	Source
FrontierMath Tiers 1–3	74.4%#22	74.0%#23	–	Epoch AI Benchmarking Hub
FrontierMath Tier 4	41.5%#31	46.3%#27	–	Epoch AI Benchmarking Hub
OTIS Mock AIME	99.2%#16	–	–	Epoch AI Benchmarking Hub
ProofBench	–	58.0%#21	55.0%#26	Vals AI
LiveBench Mathematicsnot in index	96.0%#13	–	–	LiveBench
Arena Maths	–	1504#11	–	Arena (LMArena)

Knowledge

Benchmark	high	xhigh	max	not stated	Source
LiveBench 2.0 Data Analysisnot in indexeffort not stated	–	–	–	79.6%#11	LiveBench
LiveBench Researchnot in indexeffort not stated	–	–	–	68.6%#33	LiveBench
LiveBench 1.0 Data Analysisnot in index	–	79.6%#12	–	–	LiveBench
AA-Omniscience	–	23.1#56	25#52	–	Artificial Analysis
PRBench Financeeffort not stated	–	–	–	59.5%#1	Scale AI SEAL
PRBench Legaleffort not stated	–	–	–	61.6%#1	Scale AI SEAL
Excel Modeling Benchmarknot in index	–	62.7%#31	67.4%#16	–	Vals AI
DrugDiscoveryBenchnot in index	62.2%#2	–	–	–	Scale AI SEAL
GDP.pdfnot in index	–	27.6%#12	–	27.6%#12	Epoch AI Benchmarking Hub
AA-Omniscience: accuracynot in index	–	41.5%#110	43.6%#97	–	Artificial Analysis
AA-Omniscience: non-hallucinationnot in index	–	68.5%#50	67.1%#56	–	Artificial Analysis

Instruction following

Benchmark	xhigh	max	Source
LiveBench Languagenot in index	82.8%#24	–	LiveBench
LiveBench Instruction Followingnot in index	78.0%#4	–	LiveBench
Arena Instruction Followingnot in index	–	1483#20	Arena (LMArena)

Human preference

Benchmark	max	Source
Arena Text	1494#10	Arena (LMArena)
Arena Creative Writingnot in index	1461#29	Arena (LMArena)
Arena Multi-turnnot in index	1491#20	Arena (LMArena)
Arena Longer Queriesnot in index	1496#20	Arena (LMArena)

Multimodal

Benchmark	xhigh	max	not stated	Source
Arena Vision	1302#19	1308#12	–	Arena (LMArena)
LiveBench Designnot in indexeffort not stated	–	–	74.8%#29	LiveBench
MMMU-Pro	82.0%#34	–	–	Artificial Analysis
AA-3Dnot in index	–	1037#8	–	Artificial Analysis

Long context

Benchmark	xhigh	max	Source
AA-LCR	83.0%#25	83.0%#25	Artificial Analysis
Arena Documentnot in index	–	1468#19	Arena (LMArena)
MLCRnot in index	–	43.3%#15	Artificial Analysis

Composite

Benchmark	xhigh	max	not stated	Source
Epoch Capabilities Indexnot in indexeffort not stated	–	–	157.2#23	Epoch AI Benchmarking Hub
LiveBench 2.0effort not stated	–	–	74.8%#19	LiveBench
LiveBench 1.0not in index	81.6%#7	–	–	LiveBench
AA Intelligence Index v4.3.2	45.1#37	48.1#25	–	Artificial Analysis

Where it wins

Benchmarks where this configuration ranks in the top five of every configuration measured.

#3τ³-Banking (AA)50.5% · of 202

Where to run it

Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.

Provider	Speed	First token	Input $/M	Output $/M	Context	Quantisation
Meta	127 tok/s	2.34 s	$1.25	$4.25	1.0M	–

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.150 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0018	$0.0014	1.4 min
Summarize a 30-page report	12,000 / 600	$0.018	$0.0077	1.4 min
Code edit	6,000 / 1,500	$0.014	$0.0089	1.5 min
Agentic coding session	60,000 / 4,000	$0.092	$0.043	1.9 min
Structured extraction	2,000 / 200	$0.0034	$0.0017	1.4 min