name = "Muse Spark 1.1"
description = "Muse Spark is a natively multimodal reasoning model with support for tool-use, visual chain of thought, and multi-agent orchestration."
family = "muse"
release_date = "2026-04-08"
last_updated = "2026-07-09"
attachment = true
reasoning = true
temperature = true
tool_call = true
structured_output = true
open_weights = false

[limit]
context = 1_048_576
output = 131_072

[modalities]
input = ["text", "image", "pdf", "video"]
output = ["text"]

[[benchmarks]]
name = "SWE-Bench Pro"
score = 61.5
metric = "resolve rate"
source = "https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/"
date = "2026-07-09"

[[benchmarks]]
name = "Terminal-Bench"
score = 80.0
metric = "success rate"
version = "2.1"
source = "https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/"
date = "2026-07-09"

[[benchmarks]]
name = "DeepSWE"
score = 53.3
metric = "resolve rate"
version = "1.1"
source = "https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/"
date = "2026-07-09"

[[benchmarks]]
name = "MCP Atlas"
score = 88.1
metric = "success rate"
source = "https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/"
date = "2026-07-09"

[[benchmarks]]
name = "JobBench"
score = 54.7
metric = "success rate"
source = "https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/"
date = "2026-07-09"

[[benchmarks]]
name = "Toolathlon-Verified"
score = 75.6
metric = "success rate"
source = "https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/"
date = "2026-07-09"

[[benchmarks]]
name = "Humanity's Last Exam"
score = 62.1
metric = "accuracy"
variant = "with tools"
source = "https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/"
date = "2026-07-09"

[[benchmarks]]
name = "OSWorld-Verified"
score = 80.8
metric = "success rate"
source = "https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/"
date = "2026-07-09"

[[benchmarks]]
name = "Finance Agent"
score = 57.2
metric = "accuracy"
version = "v2"
source = "https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/"
date = "2026-07-09"

[[benchmarks]]
name = "CharXiv Reasoning"
score = 88.4
metric = "accuracy"
source = "https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/"
date = "2026-07-09"

[[benchmarks]]
name = "BabyVision"
score = 76.3
metric = "accuracy"
source = "https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/"
date = "2026-07-09"