MetaReasoning modelCompare with another model

Muse Spark 1.2

Reasoning effort

Muse Spark 1.2 is a Meta proprietary reasoning model, released 5 Aug 2026. Its best configuration (xhigh reasoning effort) ranks #69 of 775 on the BenchLeader Index at 63.2 ±4.8, in the upper half. The ± is the point: 160 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 (74) and lowest in agents & tools (55). At $2.00 per million tokens blended it is pricier than most ranked models. Output speed of 315 tokens per second puts it in the fastest quarter, with a first answer in 21.9 s. It has been measured at 2 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: 315 tok/s; measured by Artificial Analysis
First answer: 22 s; first token 2.82 s
Context: 1.0M
Full answer: 23 s; median, reasoning included
Cost per run: $0.975; one full Intelligence Index run
Released: 5 Aug 2026

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

Overall index63
Reasoning70
Coding55
Agents & tools55
Maths67
Knowledge67
Human preference69
Multimodal65
Long context65
Composite74

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 1.3max	2 Sept 2026	67.2	#26
Muse Spark 1.2xhighthis page	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
xhighbest	63.2	#69	315 tok/s	22 s	$0.0018	Agents & tools 55 · Coding 55 · Composite 74 · Human preference 69 · Knowledge 67 · Long context 65 · Maths 67 · Multimodal 65 · Reasoning 70
not stated	–	–	232 tok/s	2.82 s	$0.0018	Agents & tools 45 · Composite 43 · Maths 51 · Reasoning 71

How its index has moved

8 Sept 2026 to 20 Oct 2026

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	not stated	Source
SimpleBencheffort not stated	–	74.5%#9	SimpleBench
Arena Hard Prompts	1511#17	–	Arena (LMArena)
LiveBench 2.0 Reasoningnot in indexeffort not stated	–	90.0%#13	LiveBench
LiveBench 1.0 Reasoningnot in index	90.0%#14	–	LiveBench
GPQA Diamond (AA)not in index	90.4%#62	–	Artificial Analysis
Humanity's Last Exam (AA)not in index	45.5%#49	–	Artificial Analysis
CritPt	17.7%#73	–	Artificial Analysis
Chess Puzzlesnot in index	27.0%#71	–	Epoch AI Benchmarking Hub
LMCAnot in index	48.4%#77	–	Epoch AI Benchmarking Hub
DTBenchnot in index	94.7%#52	–	Epoch AI Benchmarking Hub

Coding

Benchmark	xhigh	not stated	Source
SciCode	57.4%#28	–	SciCode
WeirdML	60.3%#50	–	WeirdML
Arena Coding	1539#11	–	Arena (LMArena)
Arena WebDev	1534#48	–	Arena (LMArena)
LiveBench 2.0 Codingnot in indexeffort not stated	–	78.2%#32	LiveBench
LiveBench 1.0 Codingnot in index	77.5%#37	–	LiveBench
SciCode (AA)not in index	57.4%#27	–	Artificial Analysis
IOI	21.8%#40	–	Vals AI
SWE-bench (Vals)not in index	86.6%#13	–	Vals AI
Code Migrationnot in index	29.9%#42	–	Vals AI
ProgramBenchnot in index	0.5%#18	–	Vals AI
Vibe Code Bench v1.1not in index	79.1%#26	–	Vals AI
DeepSWE v1.1not in index	54.9%#38	–	Epoch AI Benchmarking Hub
FrontierSWEnot in index	12.0%#19	–	Epoch AI Benchmarking Hub
Terminal-Bench 4.0 (AA)not in index	7.1%#93	–	Artificial Analysis
Terminal-Bench 2.1 (AA)not in index	80.2%#44	–	Artificial Analysis
Startrise Frontendnot in indexeffort not stated	–	70.3%#18	Startrise

Agents & tools

Benchmark	xhigh	not stated	Source
APEX-Agentseffort not stated	–	45.4%#36	Mercor
Arena Agent	-6.7#43	–	Arena (LMArena)
LiveBench Long Agenticnot in indexeffort not stated	–	21.3%#39	LiveBench
LiveBench Browsernot in indexeffort not stated	–	36.0%#34	LiveBench
LiveBench Agentic Codingnot in index	57.6%#20	–	LiveBench
Terminal-Bench 2.1 (Vals)	69.7%#28	–	Vals AI
GDPval-AA v2.1not in index	48.6%#54	–	Artificial Analysis
τ³-Banking (AA)	34.9%#49	–	Artificial Analysis
CyberBenchnot in index	69.5%#23	–	Vals AI
Finance Agent v2not in index	60.6%#4	–	Vals AI
Legal Research Benchnot in index	43.8%#20	–	Vals AI
SkillsBenchnot in index	53.0%#17	–	Vals AI
SREBenchnot in index	0.0%#31	–	Vals AI
Tax Agent Benchnot in index	32.8%#16	–	Vals AI
Terminal-Bench 4.0 (Vals)not in index	6.1%#41	–	Vals AI
AutomationBenchnot in index	40.6%#95	–	Artificial Analysis
GDP.pdf (AA)not in index	17.4%#80	–	Artificial Analysis
EnterpriseOps-Gymnot in index	47.3%#9	–	Artificial Analysis

Maths

Benchmark	xhigh	not stated	Source
ProofBencheffort not stated	–	43.0%#36	Vals AI
LiveBench Mathematicsnot in index	91.2%#31	–	LiveBench
Arena Maths	1484#34	–	Arena (LMArena)

Knowledge

Benchmark	xhigh	not stated	Source
SimpleQA Verified	60.3%#16	–	Epoch AI Benchmarking Hub
LiveBench 2.0 Data Analysisnot in indexeffort not stated	–	76.5%#22	LiveBench
LiveBench Researchnot in indexeffort not stated	–	67.7%#35	LiveBench
LiveBench 1.0 Data Analysisnot in index	76.5%#33	–	LiveBench
AA-Omniscience	27.2#42	–	Artificial Analysis
MMLU-Pro	88.3%#20	–	Vals AI
LegalBench	85.3%#23	–	Vals AI
CorpFin	70.9%#5	–	Vals AI
TaxEval	80.4%#1	–	Vals AI
BioMysteryBenchnot in index	64.8%#20	–	Vals AI
Excel Modeling Benchmarknot in index	57.0%#42	–	Vals AI
MedCodenot in index	49.4%#24	–	Vals AI
MedScribenot in index	90.1%#5	–	Vals AI
Public Benefits Benchnot in index	68.5%#10	–	Vals AI
GDP.pdfnot in index	16.0%#38	12.0%#48	Epoch AI Benchmarking Hub
AA-Omniscience: accuracynot in index	45.4%#89	–	Artificial Analysis
AA-Omniscience: non-hallucinationnot in index	66.7%#58	–	Artificial Analysis

Instruction following

Benchmark	xhigh	Source
LiveBench Languagenot in index	78.6%#38	LiveBench
LiveBench Instruction Followingnot in index	74.3%#11	LiveBench
Arena Instruction Followingnot in index	1471#36	Arena (LMArena)

Human preference

Benchmark	xhigh	Source
Arena Text	1493#12	Arena (LMArena)
Arena Creative Writingnot in index	1445#51	Arena (LMArena)
Arena Multi-turnnot in index	1514#6	Arena (LMArena)
Arena Longer Queriesnot in index	1491#27	Arena (LMArena)

Multimodal

Benchmark	xhigh	not stated	Source
Arena Vision	1304#17	–	Arena (LMArena)
LiveBench Designnot in indexeffort not stated	–	67.1%#39	LiveBench
MMMU-Pro (Vals)not in index	86.1%#22	–	Vals AI
MortgageTaxnot in index	65.4%#42	–	Vals AI
SAGEnot in index	47.7%#30	–	Vals AI

Long context

Benchmark	xhigh	Source	Trend
AA-LCR	79.0%#105	Artificial Analysis
MLCRnot in index	31.1%#20	Artificial Analysis

Composite

Benchmark	xhigh	not stated	Source
Epoch Capabilities Indexnot in indexeffort not stated	–	155.1#37	Epoch AI Benchmarking Hub
LiveBench 2.0effort not stated	–	62.4%#36	LiveBench
LiveBench 1.0not in index	78.0%#20	–	LiveBench
AA Intelligence Index v4.3.2	39.6#67	–	Artificial Analysis
Vals Multimodal Indexnot in indexeffort not stated	–	69.8%#6	Vals AI

Where it wins

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

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	243 tok/s	2.61 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	22.8 s
Summarize a 30-page report	12,000 / 600	$0.018	$0.0077	23.8 s
Code edit	6,000 / 1,500	$0.014	$0.0089	26.6 s
Agentic coding session	60,000 / 4,000	$0.092	$0.043	34.6 s
Structured extraction	2,000 / 200	$0.0034	$0.0017	22.5 s