Skip to content

Solar Open 100B VRAM requirements

How much VRAM Solar Open 100B needs
4-bit weights: 58.0 GBWith 8K context: 60.4 GBWith 32K context: 64.9 GB8-bit, 8K context: 104 GB
Parameters
103B
Active per token
12.1B
Longest context
128K
Released
Dec 2025
License
Upstage Solar License
Strongest at
Not rated

Its licence (Upstage Solar License) is not a standard open-source one: check the terms before commercial use.

How much VRAM Solar Open 100B needs

Memory Solar Open 100B needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit191 GB194 GB198 GB216 GB
8-bit102 GB104 GB109 GB127 GB
6-bit78.4 GB80.8 GB85.3 GB103 GB
5-bit68.0 GB70.4 GB74.9 GB92.9 GB
4-bit58.0 GB60.4 GB64.9 GB82.9 GB
3-bit46.7 GB49.2 GB53.7 GB71.7 GB
2-bit35.9 GB38.3 GB42.8 GB60.8 GB
1.6-bit19.7 GB22.2 GB26.7 GB44.7 GB

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

GPUs and Macs that run Solar Open 100B

The best way Solar Open 100B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: Solar Open 100B on RTX 4060 report
RTX 5060Too bigPDF: Solar Open 100B on RTX 5060 report
RTX 3070Too bigPDF: Solar Open 100B on RTX 3070 report
RTX 3080 10 GBToo bigPDF: Solar Open 100B on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: Solar Open 100B on RTX 3060 12 GB report
Arc B580Too bigPDF: Solar Open 100B on Arc B580 report
RTX 4070Too bigPDF: Solar Open 100B on RTX 4070 report
RTX 4070 SuperToo bigPDF: Solar Open 100B on RTX 4070 Super report
RTX 5070Too bigPDF: Solar Open 100B on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: Solar Open 100B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: Solar Open 100B on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: Solar Open 100B on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: Solar Open 100B on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: Solar Open 100B on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: Solar Open 100B on RTX 4080 Super report
RTX 5070 TiToo bigPDF: Solar Open 100B on RTX 5070 Ti report
RTX 5080Too bigPDF: Solar Open 100B on RTX 5080 report
Radeon RX 7900 XTXToo bigPDF: Solar Open 100B on Radeon RX 7900 XTX report
RTX 3090Too bigPDF: Solar Open 100B on RTX 3090 report
RTX 4090Too bigPDF: Solar Open 100B on RTX 4090 report
Radeon AI PRO R9700Too bigPDF: Solar Open 100B on Radeon AI PRO R9700 report
RTX 5090Too bigPDF: Solar Open 100B on RTX 5090 report
2× RTX 30902-bit40K102 tok/s1.3 sPDF: Solar Open 100B on 2× RTX 3090 report
2× RTX 40902-bit40K108 tok/s0.5 sPDF: Solar Open 100B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB2-bit44K130 tok/s1.1 sPDF: Solar Open 100B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada2-bit44K104 tok/s0.8 sPDF: Solar Open 100B on RTX 6000 Ada report
L40S2-bit44K97 tok/s1.4 sPDF: Solar Open 100B on L40S report
2× RTX 50903-bit64K132 tok/s0.4 sPDF: Solar Open 100B on 2× RTX 5090 report
4× RTX 30906-bit48K56 tok/s0.7 sPDF: Solar Open 100B on 4× RTX 3090 report
RTX PRO 6000 Blackwell6-bit60K93 tok/s0.6 sPDF: Solar Open 100B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: Solar Open 100B on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: Solar Open 100B on Mac mini M4 24 GB report
Mac mini M4 32 GBToo bigPDF: Solar Open 100B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: Solar Open 100B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GBToo bigPDF: Solar Open 100B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB2-bit44K37 tok/s25 sPDF: Solar Open 100B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB6-bit64K32 tok/s12 sPDF: Solar Open 100B on M4 Max 128 GB report
M5 Max 128 GB6-bit64K35 tok/s3.6 sPDF: Solar Open 100B on M5 Max 128 GB report
M3 Ultra 96 GB4-bit48K49 tok/s6.9 sPDF: Solar Open 100B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit128K32 tok/s6.9 sPDF: Solar Open 100B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit128K32 tok/s6.9 sPDF: Solar Open 100B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit128K43 tok/s2.2 sPDF: Solar Open 100B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB2-bit44K30 tok/s7.1 sPDF: Solar Open 100B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB6-bit64K15 tok/s7.1 sPDF: Solar Open 100B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit60K16 tok/s2.5 sPDF: Solar Open 100B on RTX Spark report
DGX Spark8-bit60K14 tok/s2.5 sPDF: Solar Open 100B on DGX Spark report
Dual DGX Spark8-bit128K14 tok/s1.4 sPDF: Solar Open 100B on Dual DGX Spark report
DGX Station GB3008-bit128K183 tok/s0.2 sPDF: Solar Open 100B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit128K76 tok/s0.3 sPDF: Solar Open 100B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit128K76 tok/s0.2 sPDF: Solar Open 100B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit128K76 tok/s0.1 sPDF: Solar Open 100B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM5-bit36K113 tok/s1.6 sPDF: Solar Open 100B on A100 80 GB SXM report
H100 SXM5-bit36K153 tok/s0.5 sPDF: Solar Open 100B on H100 SXM report
4× H200 NVL8-bit128K149 tok/s0.2 sPDF: Solar Open 100B on 4× H200 NVL report
8× H200 NVL8-bit128K149 tok/s0.1 sPDF: Solar Open 100B 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 mradermacher/Solar-Open-100B-GGUF
vLLM
vllm serve upstage/Solar-Open-100B
MLX (Mac)
mlx_lm.chat --model mlx-community/Solar-Open-100B-8bit

Questions

How much VRAM does Solar Open 100B need?

Solar Open 100B needs 58.0 GB for its weights at 4-bit, 60.4 GB with an 8K context and 64.9 GB with 32K. At 8-bit, 104 GB.

What is the minimum VRAM for Solar Open 100B?

About 38.3 GB at 2-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 2× RTX 3090.

How big is Solar Open 100B?

103B parameters (12.1B active per token): 191 GB of weights at 16-bit, 58.0 GB at 4-bit.

Can Solar Open 100B run on an RTX 4090?

Not on one card. The smallest machine here that runs it is the 2× RTX 3090.

More models