Compare Models

Muse Spark 1.3

Muse Spark 1.3 is a model from Meta, released September 2, 2026. It ranks #22 of 50 models on the Vals Index with 53.20%. Its best result is #3 of 81 on Harvey's Legal Agent Benchmark.

Release Date: Sep 2, 2026

Developer Meta  🇺🇸
Context Window 1M
Max Output Tokens 131k
Token Costs (in/out) $1.25/4.25
Weights Private
Input Modalities

Accuracy

53.20 % ± 1.07

Cost / Test (Vals Index)

$ 3.376

Latency

32 min 46 s

Fallback Rate 0.00%
Refusal Rate 0.13%
Vals Index
BenchmarksAccuracyRankings
Vals Index

53.20%

±1.07
22/50
Code Migration

27.58%

±4.07
45/79
CyberBench v1.1

69.41%

±5.76
26/47
EMB

62.71%

±2.96
31/76
Finance Agent (v2)

58.90%

±0.44
8/80
Legal Research Bench

40.87%

±3.42
28/80
ProofBench v1.1

55.00%

±5.00
27/50
Tax Agent Bench

35.97%

±3.42
12/71
Vibe Code Bench v1.1

82.86%

±2.90
20/114
Harvey's Legal Agent Benchmark

22.92%

±3.42
3/81
IOI

43.94%

±1.45
33/44
ProgramBench

0.50%

±0.50
22/64
Terminal-Bench 4.0

10.61%

±1.75
38/50
Terminal-Bench-Science

4.29%

±2.44
25/41
Proprietary BenchmarksAcademic BenchmarksIndustry Partners
Vals
Default Provider : Meta
Temperature: 1
Top P: Default
Top K: Default
Max Output Tokens: 131,072
Reasoning Effort: xhigh

Compared with

Muse Spark 1.3 vs MiMo V2.6 Flash 14 shared benchmarks Muse Spark 1.3 vs GPT-5.6 Terra 14 shared benchmarks Muse Spark 1.3 vs GLM 5.3 14 shared benchmarks Muse Spark 1.3 vs Grok 4.6 14 shared benchmarks Muse Spark 1.3 vs Claude Haiku 5.5 14 shared benchmarks Muse Spark 1.3 vs Claude Sonnet 5 14 shared benchmarks Muse Spark 1.3 vs Grok 4.7 14 shared benchmarks Muse Spark 1.3 vs GPT-5.6 Luna 14 shared benchmarks Muse Spark 1.3 vs Gemini 3.8 Flash 14 shared benchmarks Muse Spark 1.3 vs DeepSeek V4.1 Flash 14 shared benchmarks