Solar Open 2 — which GPUs fit

Upstage · 250.0B (active 15.0B) · MoE · ctx 1024K · upstage-solar · HuggingFace

Solar Open 2 (Upstage) is a 250.0B parameters MoE (15.0B active per token) open-weight model tracked on AI Hashrate. At 8K context we estimate roughly 138.88 GB VRAM for Q4 (weights 137.5 GB + KV 0.38 GB + 1 GB overhead) and 501.38 GB for FP16. About 5 GPUs/accelerators in our catalog fully fit this model at Q4 under that context assumption. Fastest Q4 configs: B200 192GB SXM ≈ 510.3 tok/s (estimated); MI300X 192GB ≈ 333.6 tok/s (estimated); H200 141GB SXM5 ≈ 306.2 tok/s (estimated); DGX Spark (GH200 480GB) ≈ 30.9 tok/s (estimated); Mac Studio M2 Ultra 192GB ≈ 30.0 tok/s (estimated). Use the table below for VRAM fit and tok/s per dollar (list MSRP). Estimates follow memory-bandwidth math; measured rows override when present. See methodology for details. Methodology.

25 hardware configs — sorted by tok/s. Default fit context 8K.

Hardware	VRAM	Quant	tok/s	Fits?	tok/s/$
B200 192GB SXM	192	Q4	510.3 est.	Yes	0.013
Gaudi 3 128GB	128	Q4	334.8 est.	No	0.027
MI300X 192GB	192	Q4	333.6 est.	Yes	0.022
H200 141GB SXM5	141	Q4	306.2 est.	Tight	0.009
H100 80GB SXM5	80	Q4	134.7 est.	No	0.004
A100 80GB SXM4	80	Q4	82.0 est.	No	0.005
B200 192GB SXM	192	FP16	39.1 est.	No	0.001
MacBook Pro M4 Max 128GB	128	Q4	31.8 est.	No	0.007
DGX Spark (GH200 480GB)	480	Q4	30.9 est.	Yes	0.01
Mac Studio M2 Ultra 192GB	192	Q4	30.0 est.	Tight	0.004
Mac Studio M3 Ultra 96GB	96	Q4	29.0 est.	No	0.007
MI300X 192GB	192	FP16	25.9 est.	No	0.002
MacBook Pro M3 Max 128GB	128	Q4	23.3 est.	No	0.006
RTX 6000 Ada 48GB	48	Q4	13.9 est.	No	0.002
H200 141GB SXM5	141	FP16	12.6 est.	No	0.0
L40S 48GB	48	Q4	12.5 est.	No	0.002
RTX A6000 48GB	48	Q4	11.1 est.	No	0.002
DGX Spark (GH200 480GB)	480	FP16	9.8 est.	No	0.003
Gaudi 3 128GB	128	FP16	7.1 est.	No	0.001
Ryzen AI Max+ 395 128GB	96	Q4	7.0 est.	No	0.004
MacBook Pro M1 Max 64GB	64	Q4	5.8 est.	No	0.001
Mac mini M4 Pro 64GB	64	Q4	3.5 est.	No	0.002
Mac Studio M2 Ultra 192GB	192	FP16	2.4 est.	No	0.0
Ryzen AI Max+ 395 64GB	48	Q4	1.7 est.	No	0.001
Ryzen AI Max 390 64GB	48	Q4	1.7 est.	No	0.001

Fit checks

RTX 6000 Ada 48GB — does not fit · ≈13.9 tok/s Q4 est.
L40S 48GB — does not fit · ≈12.5 tok/s Q4 est.
Mac Studio M3 Ultra 96GB — does not fit · ≈29.0 tok/s Q4 est.
Mac Studio M2 Ultra 192GB — fits · ≈30.0 tok/s Q4 est.

12 popular retail GPUs; every row in the table above also links its Fits? verdict to the full fit check.

Same family: Upstage

All Upstage models →