Meta
Muse Wonder
Muse Wonder is a frontier multimodal reasoning model from Meta for personal and professional agents, long-horizon coding, and computer use. It combines Active Memory, which carries relevant context across conversations, with Active Focus, which keeps it working through blockers toward a goal.
meta/muse-wonder-1.0
Context
1M
Output
131.1K
Released
2026-10-12
Knowledge cutoff
—
Types
Reasoning
Tool use
Structured output
Attachments
Open weights
Provider pricing
- Price
- $4.00 / $20.00
- Context
- 1M
- Status
- Stable
- Price
- $4.00 / $20.00
- Context
- 1M
- Status
- Stable
- Price
- $4.00 / $20.00
- Context
- 1M
- Status
- Stable
- Price
- $4.00 / $20.00
- Context
- 1M
- Status
- Stable
Benchmarks
Harvey Legal Agent Benchmark
10 · all-pass rate
Finance Agent
56.6 · accuracy
HealthBench Professional
60.5 · length-normalized weighted rubric score
AutomationBench
39.1 · strict pass@1
GDP.pdf
30.8 · score
JobBench
70.5 · mean reward
OSWorld
72.6 · mean partial score
OSWorld
34.3 · binary completion rate
Terminal-Bench
52 · accuracy
DeepSWE
78.8 · pass@1
SWE-Marathon
53.1 · pass@1
ProgramBench
50 · share of rollouts passing at least 95% of tests
Artificial Analysis Intelligence Index
52 · index score
From the same lab
Muse Spark Open
meta/muse-spark-open
Muse Wonder Full
meta/muse-wonder-full
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