Skip to content

Muse Spark Open 816B A42B VRAM requirements

How much VRAM Muse Spark Open 816B A42B needs
4-bit weights: 464 GBWith 8K context: 466 GBWith 32K context: 468 GB8-bit, 8K context: 815 GB
Parameters
822B
Active per token
42B
Longest context
1M
Released
Sep 2026
License
Muse Model License
Strongest at
Chat, Docs

How much VRAM Muse Spark Open 816B A42B needs

Memory Muse Spark Open 816B A42B needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit1,531 GB1,533 GB1,535 GB1,541 GB
8-bit813 GB815 GB817 GB823 GB
6-bit628 GB630 GB631 GB637 GB
5-bit545 GB546 GB548 GB554 GB
4-bit464 GB466 GB468 GB474 GB
3-bit374 GB376 GB378 GB384 GB
2-bit287 GB289 GB291 GB297 GB
1.6-bit158 GB160 GB161 GB167 GB

One conversation, 16-bit context cache, runtime overhead included.

GPUs and Macs that run Muse Spark Open 816B A42B

The best way Muse Spark Open 816B A42B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: Muse Spark Open 816B A42B on RTX 4060 report
RTX 5060Too bigPDF: Muse Spark Open 816B A42B on RTX 5060 report
RTX 3070Too bigPDF: Muse Spark Open 816B A42B on RTX 3070 report
RTX 3080 10 GBToo bigPDF: Muse Spark Open 816B A42B on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: Muse Spark Open 816B A42B on RTX 3060 12 GB report
Arc B580Too bigPDF: Muse Spark Open 816B A42B on Arc B580 report
RTX 4070Too bigPDF: Muse Spark Open 816B A42B on RTX 4070 report
RTX 4070 SuperToo bigPDF: Muse Spark Open 816B A42B on RTX 4070 Super report
RTX 5070Too bigPDF: Muse Spark Open 816B A42B on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: Muse Spark Open 816B A42B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: Muse Spark Open 816B A42B on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: Muse Spark Open 816B A42B on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: Muse Spark Open 816B A42B on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: Muse Spark Open 816B A42B on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: Muse Spark Open 816B A42B on RTX 4080 Super report
RTX 5070 TiToo bigPDF: Muse Spark Open 816B A42B on RTX 5070 Ti report
RTX 5080Too bigPDF: Muse Spark Open 816B A42B on RTX 5080 report
Radeon RX 7900 XTXToo bigPDF: Muse Spark Open 816B A42B on Radeon RX 7900 XTX report
RTX 3090Too bigPDF: Muse Spark Open 816B A42B on RTX 3090 report
RTX 4090Too bigPDF: Muse Spark Open 816B A42B on RTX 4090 report
Radeon AI PRO R9700Too bigPDF: Muse Spark Open 816B A42B on Radeon AI PRO R9700 report
RTX 5090Too bigPDF: Muse Spark Open 816B A42B on RTX 5090 report
2× RTX 3090Too bigPDF: Muse Spark Open 816B A42B on 2× RTX 3090 report
2× RTX 4090Too bigPDF: Muse Spark Open 816B A42B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GBToo bigPDF: Muse Spark Open 816B A42B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 AdaToo bigPDF: Muse Spark Open 816B A42B on RTX 6000 Ada report
L40SToo bigPDF: Muse Spark Open 816B A42B on L40S report
2× RTX 5090Too bigPDF: Muse Spark Open 816B A42B on 2× RTX 5090 report
4× RTX 3090Too bigPDF: Muse Spark Open 816B A42B on 4× RTX 3090 report
RTX PRO 6000 BlackwellToo bigPDF: Muse Spark Open 816B A42B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: Muse Spark Open 816B A42B on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: Muse Spark Open 816B A42B on Mac mini M4 24 GB report
Mac mini M4 32 GBToo bigPDF: Muse Spark Open 816B A42B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: Muse Spark Open 816B A42B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GBToo bigPDF: Muse Spark Open 816B A42B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GBToo bigPDF: Muse Spark Open 816B A42B on Mac mini M4 Pro 64 GB report
M4 Max 128 GBToo bigPDF: Muse Spark Open 816B A42B on M4 Max 128 GB report
M5 Max 128 GBToo bigPDF: Muse Spark Open 816B A42B on M5 Max 128 GB report
M3 Ultra 96 GBToo bigPDF: Muse Spark Open 816B A42B on M3 Ultra 96 GB report
M3 Ultra 256 GB1.6-bit352K40 tok/s26 sPDF: Muse Spark Open 816B A42B on M3 Ultra 256 GB report
M3 Ultra 512 GB2-bit1M26 tok/s26 sPDF: Muse Spark Open 816B A42B on M3 Ultra 512 GB report
M5 Ultra 512 GB2-bit1M35 tok/s8.1 sPDF: Muse Spark Open 816B A42B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GBToo bigPDF: Muse Spark Open 816B A42B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GBToo bigPDF: Muse Spark Open 816B A42B on Ryzen AI Max+ 395 128 GB report
RTX SparkToo bigPDF: Muse Spark Open 816B A42B on RTX Spark report
DGX SparkToo bigPDF: Muse Spark Open 816B A42B on DGX Spark report
Dual DGX Spark1.6-bit1M20 tok/s5.3 sPDF: Muse Spark Open 816B A42B on Dual DGX Spark report
DGX Station GB3001.6-bit1M183 tok/s0.7 sPDF: Muse Spark Open 816B A42B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell1.6-bit352K94 tok/s1.3 sPDF: Muse Spark Open 816B A42B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell2-bit1M61 tok/s0.7 sPDF: Muse Spark Open 816B A42B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell6-bit1M32 tok/s0.4 sPDF: Muse Spark Open 816B A42B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXMToo bigPDF: Muse Spark Open 816B A42B on A100 80 GB SXM report
H100 SXMToo bigPDF: Muse Spark Open 816B A42B on H100 SXM report
4× H200 NVL4-bit1M89 tok/s0.7 sPDF: Muse Spark Open 816B A42B on 4× H200 NVL report
8× H200 NVL8-bit1M59 tok/s0.4 sPDF: Muse Spark Open 816B A42B on 8× H200 NVL report

Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.

How to run it

llama.cpp
llama-server -hf unsloth/Muse-Spark-Open-816B-A42B-GGUF
vLLM
vllm serve meta-models/Muse-Spark-Open-816B-A42B
MLX (Mac)
mlx_lm.chat --model mlx-community/Muse-Spark-Open-816B-A42B-4bit

Questions

How much VRAM does Muse Spark Open 816B A42B need?

Muse Spark Open 816B A42B needs 464 GB for its weights at 4-bit, 466 GB with an 8K context and 468 GB with 32K. At 8-bit, 815 GB.

What is the minimum VRAM for Muse Spark Open 816B A42B?

About 160 GB at 1.6-bit with an 8K context, the lowest precision we recommend for a model this size. The smallest machine here that runs it without spilling is the M3 Ultra 256 GB.

How big is Muse Spark Open 816B A42B?

822B parameters (42B active per token): 1,531 GB of weights at 16-bit, 464 GB at 4-bit.

Can Muse Spark Open 816B A42B run on an RTX 4090?

Not on one card. The smallest machine here that runs it is the M3 Ultra 256 GB.

More models