Muse Spark Open 816B A42B
Parameters
816B
Context Length
-
Modality
Text
Architecture
-
License
other
Release Date
15 Sept 2026
Knowledge Cutoff
-
API Pricing (per 1M)
Self-hosted only
System Requirements
VRAM requirements for different quantization methods and context sizes
1,024 tokens
Consumer
100x RTX 4090
24GB VRAM
Datacenter
25x NVIDIA A100
80GB VRAM
Apple Silicon
21x Apple M3 Max
128GB VRAM
131,072 tokens
Consumer
101x RTX 4090
24GB VRAM
Datacenter
26x NVIDIA A100
80GB VRAM
Apple Silicon
22x Apple M3 Max
128GB VRAM
Architecture Diagram
Evaluation Benchmarks
Rank
#167
| Benchmark | Score | Rank |
|---|---|---|
Coding | 0.74 | 48 |
General | 0.58 | 57 |
Data Analysis | 0.60 | 57 |
LiveBench Average | 0.58 | 57 |
Reasoning | 0.74 | 59 |
Rankings
Overall Rank
#167
Coding Rank
#92
About Muse Spark Open 816B A42B
Muse Spark Open 816B A42B is a massive mixture-of-experts open foundation model developed by Meta. It activates 42B parameters per token to deliver frontier reasoning and multimodal language performance.
Technical Specifications
Resources
About Muse Spark
The Muse Spark model family developed by Meta.