Laguna S 2.1 — which GPUs fit

Poolside · 118.0B (active 8.0B) · MoE · ctx 1024K · openmdw-1.1 · HuggingFace

Laguna S 2.1 (Poolside) is a 118.0B parameters MoE (8.0B active per token) open-weight model tracked on AI Hashrate. At 8K context we estimate roughly 67.4 GB VRAM for Q4 (weights 64.9 GB + KV 1.5 GB + 1 GB overhead) and 238.5 GB for FP16. About 12 GPUs/accelerators in our catalog fully fit this model at Q4 under that context assumption. Fastest Q4 configs: B200 192GB SXM ≈ 956.7 tok/s (estimated); MI300X 192GB ≈ 625.5 tok/s (estimated); H200 141GB SXM5 ≈ 574.0 tok/s (estimated); H100 80GB SXM5 ≈ 400.6 tok/s (estimated); Gaudi 3 128GB ≈ 386.6 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.

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

Hardware	VRAM	Quant	tok/s	Fits?	tok/s/$
B200 192GB SXM	192	Q4	956.7 est.	Yes	0.024
MI300X 192GB	192	Q4	625.5 est.	Yes	0.042
H200 141GB SXM5	141	Q4	574.0 est.	Yes	0.016
H100 80GB SXM5	80	Q4	400.6 est.	Yes	0.013
Gaudi 3 128GB	128	Q4	386.6 est.	Yes	0.031
B200 192GB SXM	192	FP16	324.0 est.	No	0.008
A100 80GB SXM4	80	Q4	243.8 est.	Yes	0.016
MI300X 192GB	192	FP16	214.6 est.	No	0.014
RTX 6000 Ada 48GB	48	Q4	110.5 est.	No	0.016
H200 141GB SXM5	141	FP16	104.8 est.	No	0.003
L40S 48GB	48	Q4	99.5 est.	No	0.012
MacBook Pro M4 Max 128GB	128	Q4	96.2 est.	Yes	0.022
RTX 5090 32GB	32	Q4	91.7 est.	No	0.046
RTX A6000 48GB	48	Q4	88.4 est.	No	0.02
MacBook Pro M3 Max 128GB	128	Q4	70.5 est.	Yes	0.017
Gaudi 3 128GB	128	FP16	58.7 est.	No	0.005
DGX Spark (GH200 480GB)	480	Q4	58.0 est.	Yes	0.019
Mac Studio M3 Ultra 96GB	96	Q4	57.7 est.	Yes	0.014
Mac Studio M2 Ultra 192GB	192	Q4	56.3 est.	Yes	0.008
MacBook Pro M1 Max 64GB	64	Q4	46.8 est.	No	0.012
V100 32GB SXM2	32	Q4	46.0 est.	No	0.005
Arc Pro B70 32GB	32	Q4	30.6 est.	No	0.031
RTX 4090 24GB	24	Q4	29.0 est.	No	0.018
RTX 4090D 24GB (China)	24	Q4	29.0 est.	No	0.018
RTX 3090 Ti 24GB	24	Q4	29.0 est.	No	0.015
Mac mini M4 Pro 64GB	64	Q4	27.8 est.	No	0.013
RX 7900 XTX 24GB	24	Q4	27.6 est.	No	0.028
RTX 3090 24GB	24	Q4	26.9 est.	No	0.018
Ryzen AI Max+ 395 128GB	96	Q4	25.4 est.	Tight	0.013
H100 80GB SXM5	80	FP16	23.5 est.	No	0.001
Mac Studio M2 Ultra 192GB	192	FP16	19.8 est.	No	0.003
RX 7900 XT 20GB	20	Q4	16.0 est.	No	0.018
DGX Spark (GH200 480GB)	480	FP16	15.9 est.	Yes	0.005
A100 80GB SXM4	80	FP16	14.3 est.	No	0.001
Ryzen AI Max+ 395 64GB	48	Q4	13.9 est.	No	0.009
Ryzen AI Max 390 64GB	48	Q4	13.9 est.	No	0.011
MacBook Pro M1 Max 32GB	32	Q4	11.7 est.	No	0.004
MacBook Pro M4 Max 128GB	128	FP16	5.5 est.	No	0.001
Mac Studio M3 Ultra 96GB	96	FP16	5.1 est.	No	0.001
MacBook Pro M3 Max 128GB	128	FP16	4.1 est.	No	0.001
Mac mini M4 32GB	32	Q4	3.1 est.	No	0.003
Ryzen AI Max+ 395 128GB	96	FP16	1.2 est.	No	0.001

Fit checks

RTX 5090 32GB — does not fit · ≈91.7 tok/s Q4 est.
RTX 3090 Ti 24GB — does not fit · ≈29.0 tok/s Q4 est.
RTX 4090 24GB — does not fit · ≈29.0 tok/s Q4 est.
RTX 6000 Ada 48GB — does not fit · ≈110.5 tok/s Q4 est.
RX 7900 XTX 24GB — does not fit · ≈27.6 tok/s Q4 est.
RTX 3090 24GB — does not fit · ≈26.9 tok/s Q4 est.
L40S 48GB — does not fit · ≈99.5 tok/s Q4 est.
Mac Studio M3 Ultra 96GB — fits · ≈57.7 tok/s Q4 est.
Mac Studio M2 Ultra 192GB — fits · ≈56.3 tok/s Q4 est.
RX 7900 XT 20GB — does not fit · ≈16.0 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: Poolside

All Poolside models →