Compare Models
Muse Spark 1.1

Release Date: Jul 9, 2026

Developer Meta 🇺🇸
Context Window 1M
Max Output Tokens 256k
Token Costs (in/out) $1.25/4.25
Weights Private
Input Modalities

Accuracy

54.75 % ± 1.15

Cost / Test (Vals Index)

$ 1.251

Latency

13 min 48 s

Vals Index
BenchmarksAccuracyRankings
Vals Index

0.0%

±1.15
15/46
Code Migration

0.0%

±4.07
18/49
EMB

0.0%

±3.15
22/46
Finance Agent (v2)

0.0%

±0.79
5/49
Time Horizon Index: KSP

0.0%

±0.00
7/7
Legal Research Bench

0.0%

±3.37
15/49
MedScribe

0.0%

±1.95
3/84
MortgageTax

0.0%

±0.93
36/95
SAGE

0.0%

±3.28
29/76
Public Benefits Bench v1.1

0.0%

±1.27
15/30
TaxEval v2

0.0%

±0.77
2/140
Vibe Code Bench v1.1

0.0%

±3.33
13/84
Harvey's Legal Agent Benchmark

0.0%

±3.22
2/50
GPQA Diamond

0.0%

±2.10
21/133
LiveCodeBench

0.0%

±0.99
27/138
LegalBench

0.0%

±0.42
22/137
MMLU Pro

0.0%

±0.32
15/133
MMMU

0.0%

±0.82
16/89
ProgramBench

0.0%

±0.00
22/39
SkillsBench

0.0%

±4.44
9/28
SWE-bench

0.0%

±1.72
17/83
Terminal-Bench 2.1

0.0%

±0.99
17/54
Proprietary BenchmarksAcademic BenchmarksIndustry Partners
Vals
Default Provider : Meta
Temperature: Default
Top P: Default
Top K: Default
Max Output Tokens: 256,000
Reasoning Effort: xhigh

Updates

Jul 9, 2026

We evaluated Meta’s new Muse Spark 1.1 across our benchmark suite.

  • It debuts at #14 on the Vals Index (54.75%), narrowly behind GPT 5.5 (57.41%). At $1.25/test, it is the fourth-lowest-cost model in the top 20; it also averaged 827.8 seconds on the index, roughly 2–4× faster than the top three models.

  • It takes #2 on Finance Agent v2 (57.21%), just 0.65 points behind Gemini 3.5 Flash.

  • The largest generational gain is on Vibe Code Bench. Compared with Muse Spark, Muse Spark 1.1 rises from #43 to #5 and improves by 52.48 points (19.67% → 72.16%).

  • Muse Spark 1.1 sets new highs on domain-specific work: #1 on MedScribe (88.89%), #1 on TaxEval v2 (79.72%), and #1 on Harvey’s Legal Agent Benchmark (20.00%). On MedScribe, it is nearly twice as fast as the #2 model.

The model has a 1M-token context window and supports up to 256k output tokens. Evaluations were run with reasoning effort set to “xhigh” and default temperature and top-p.

Congrats to the Meta team on the release!