Meta
Muse Wonder Full
Muse Wonder Full is the most capable model in Meta's Muse Wonder series, built for long-running professional agents and ultra-long-horizon software engineering. It shares Muse Wonder's Active Memory and Active Focus capabilities and generally scores higher on Meta's agentic and coding evaluations.
meta/muse-wonder-full
上下文
1M
输出
—
发布
2026-10-12
知识截止
—
种类
推理
工具调用
结构化输出
附件
开放权重
服务商报价
暂无服务商上架该模型。
评测
Harvey Legal Agent Benchmark
13.1 · all-pass rate
Finance Agent
59.1 · accuracy
HealthBench Professional
62.3 · length-normalized weighted rubric score
AutomationBench
45.4 · strict pass@1
GDP.pdf
33.2 · score
JobBench
75.3 · mean reward
OSWorld
78.4 · mean partial score
OSWorld
39.4 · binary completion rate
Terminal-Bench
57.1 · accuracy
DeepSWE
79 · pass@1
SWE-Marathon
57.5 · pass@1
ProgramBench
59 · share of rollouts passing at least 95% of tests
Artificial Analysis Intelligence Index
56 · index score
同一实验室
Muse Spark Open
meta/muse-spark-open
Muse Wonder
meta/muse-wonder-1.0
Muse Spark 1.3
meta/muse-spark-1.3
Muse Glimmer 30B
meta/muse-glimmer-30b
Muse Spark 1.2
meta/muse-spark-1.2
Muse Spark 1.1
meta/muse-spark-1.1
Llama 4 Maverick 17B Instruct
meta/llama-4-maverick-17b-instruct
Llama 4 Scout 17B Instruct
meta/llama-4-scout-17b-instruct