Qwen2.5-7B-Instruct — which GPUs fit

Alibaba · 7.6B · ctx 128K · apache-2.0 · HuggingFace

Qwen2.5-7B-Instruct (Alibaba) is a 7.6B parameters open-weight model tracked on AI Hashrate. At 8K context we estimate roughly 6.7 GB VRAM for Q4 (weights 4.18 GB + KV 1.52 GB + 1 GB overhead) and 17.72 GB for FP16. About 50 GPUs/accelerators in our catalog fully fit this model at Q4 under that context assumption. Fastest Q4 configs: B200 192GB SXM ≈ 669.9 tok/s (estimated); MI300X 192GB ≈ 443.8 tok/s (estimated); H200 141GB SXM5 ≈ 401.9 tok/s (estimated); H100 80GB SXM5 ≈ 280.5 tok/s (estimated); Gaudi 3 128GB ≈ 274.3 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.

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

Hardware	VRAM	Quant	tok/s	Fits?	tok/s/$
B200 192GB SXM	192	Q4	669.9 est.	Yes	0.017
MI300X 192GB	192	Q4	443.8 est.	Yes	0.03
H200 141GB SXM5	141	Q4	401.9 est.	Yes	0.011
H100 80GB SXM5	80	Q4	280.5 est.	Yes	0.009
Gaudi 3 128GB	128	Q4	274.3 est.	Yes	0.022
B200 192GB SXM	192	FP16	184.2 est.	Yes	0.005
A100 80GB SXM4	80	Q4	154.05 measured	Yes	0.01
RTX 5090 32GB	32	Q4	150.0 est.	Yes	0.075
MI300X 192GB	192	FP16	122.0 est.	Yes	0.008
H200 141GB SXM5	141	FP16	110.5 est.	Yes	0.003
RTX 4090 24GB	24	Q4	84.4 est.	Yes	0.053
RTX 4090D 24GB (China)	24	Q4	84.4 est.	Yes	0.053
RTX 3090 Ti 24GB	24	Q4	84.4 est.	Yes	0.042
A100 80GB SXM4	80	FP16	84.28 measured	Yes	0.006
RTX 5080 16GB	16	Q4	80.4 est.	Yes	0.08
RX 7900 XTX 24GB	24	Q4	80.4 est.	Yes	0.08
RTX 6000 Ada 48GB	48	Q4	80.4 est.	Yes	0.012
RTX 3090 24GB	24	Q4	78.4 est.	Yes	0.052
H100 80GB SXM5	80	FP16	77.1 est.	Yes	0.003
Gaudi 3 128GB	128	FP16	75.4 est.	Yes	0.006
RTX 5070 Ti 16GB	16	Q4	75.0 est.	Yes	0.1
L40S 48GB	48	Q4	72.3 est.	Yes	0.009
Mac Studio M3 Ultra 96GB	96	Q4	68.6 est.	Yes	0.017
RX 7900 XT 20GB	20	Q4	67.0 est.	Yes	0.075
RTX A6000 48GB	48	Q4	64.3 est.	Yes	0.014
RTX 3080 10GB	10	Q4	63.6 est.	Yes	0.091
RTX 4080 Super 16GB	16	Q4	61.6 est.	Yes	0.062
RTX 4070 Ti Super 16GB	16	Q4	56.3 est.	Yes	0.07
RTX 5070 12GB	12	Q4	56.3 est.	Yes	0.103
RX 7900 XTX 24GB	24	FP16	56.0 measured	Yes	0.056
RX 9070 16GB	16	Q4	53.6 est.	Yes	0.098
RX 9070 XT 16GB	16	Q4	53.6 est.	Yes	0.089
RX 7800 XT 16GB	16	Q4	52.2 est.	Yes	0.105
RTX 5080 16GB	16	FP16	51.5 est.	No	0.052
RX 7900 GRE 16GB	16	Q4	48.2 est.	Yes	0.088
RTX 5070 Ti 16GB	16	FP16	48.1 est.	No	0.064
Arc A770 16GB	16	Q4	46.9 est.	Yes	0.134
RTX 4070 Ti 12GB	12	Q4	46.4 est.	Yes	0.058
MacBook Pro M4 Max 128GB	128	Q4	45.7 est.	Yes	0.01
RTX 4070 12GB	12	Q4	42.2 est.	Yes	0.07
RTX 4070 Super 12GB	12	Q4	42.2 est.	Yes	0.07
RTX 5090 32GB	32	FP16	41.3 est.	Yes	0.021
RTX 4080 Super 16GB	16	FP16	39.5 est.	No	0.04
Arc B580 12GB	12	Q4	38.2 est.	Yes	0.153
RTX 5060 8GB	8	Q4	37.5 est.	Yes	0.125
RTX 5060 Ti 8GB	8	Q4	37.5 est.	Yes	0.099
RTX 5060 Ti 16GB	16	Q4	37.5 est.	Yes	0.087
RTX 3070 8GB	8	Q4	37.5 est.	Yes	0.075
RX 7700 XT 12GB	12	Q4	36.2 est.	Yes	0.081
RTX 4070 Ti Super 16GB	16	FP16	36.0 est.	No	0.045
RX 9070 16GB	16	FP16	34.3 est.	No	0.062
RX 9070 XT 16GB	16	FP16	34.3 est.	No	0.057
MacBook Pro M3 Max 128GB	128	Q4	33.5 est.	Yes	0.008
RX 7800 XT 16GB	16	FP16	33.5 est.	No	0.067
Arc B570 10GB	10	Q4	31.8 est.	Yes	0.145
RX 7900 GRE 16GB	16	FP16	30.9 est.	No	0.056
RTX 3060 12GB	12	Q4	30.1 est.	Yes	0.091
Arc A770 16GB	16	FP16	30.0 est.	No	0.086
Mac Studio M2 Ultra 192GB	192	Q4	28.0 measured	Yes	0.004
T4 16GB	16	Q4	26.8 est.	Yes	0.013
RTX 4060 Ti 8GB	8	Q4	24.1 est.	Yes	0.06
RTX 4060 Ti 16GB	16	Q4	24.1 est.	Yes	0.048
RX 7600 8GB	8	Q4	24.1 est.	Yes	0.09
RX 7600 XT 16GB	16	Q4	24.1 est.	Yes	0.073
RTX 5060 Ti 16GB	16	FP16	24.0 est.	No	0.056
RTX 4090 24GB	24	FP16	23.2 est.	Yes	0.015
RTX 4090D 24GB (China)	24	FP16	23.2 est.	Yes	0.015
RTX 3090 Ti 24GB	24	FP16	23.2 est.	Yes	0.012
Mac mini M4 Pro 64GB	64	Q4	22.9 est.	Yes	0.01
RTX 4060 8GB	8	Q4	22.8 est.	Yes	0.076
RTX 6000 Ada 48GB	48	FP16	22.1 est.	Yes	0.003
RTX 3090 24GB	24	FP16	21.6 est.	Yes	0.014
RTX 5070 12GB	12	FP16	20.3 est.	No	0.037
L40S 48GB	48	FP16	19.9 est.	Yes	0.002
Mac Studio M3 Ultra 96GB	96	FP16	18.9 est.	Yes	0.005
Mac Studio M2 Ultra 192GB	192	FP16	18.4 est.	Yes	0.003
RX 7900 XT 20GB	20	FP16	18.4 est.	Yes	0.02
RTX A6000 48GB	48	FP16	17.7 est.	Yes	0.004
T4 16GB	16	FP16	17.2 est.	No	0.009
RTX 4070 Ti 12GB	12	FP16	16.7 est.	No	0.021
RTX 3080 10GB	10	FP16	15.9 est.	No	0.023
RTX 4060 Ti 16GB	16	FP16	15.4 est.	No	0.031
RX 7600 XT 16GB	16	FP16	15.4 est.	No	0.047
RTX 4070 12GB	12	FP16	15.2 est.	No	0.025
RTX 4070 Super 12GB	12	FP16	15.2 est.	No	0.025
Arc B580 12GB	12	FP16	13.8 est.	No	0.055
RX 7700 XT 12GB	12	FP16	13.0 est.	No	0.029
MacBook Pro M4 Max 128GB	128	FP16	12.6 est.	Yes	0.003
RTX 3060 12GB	12	FP16	10.9 est.	No	0.033
Mac mini M4 32GB	32	Q4	10.0 est.	Yes	0.01
MacBook Pro M3 Max 128GB	128	FP16	9.2 est.	Yes	0.002
Arc B570 10GB	10	FP16	8.0 est.	No	0.037
Mac mini M4 Pro 64GB	64	FP16	6.3 est.	Yes	0.003
RTX 5060 8GB	8	FP16	6.0 est.	No	0.02
RTX 5060 Ti 8GB	8	FP16	6.0 est.	No	0.016
RTX 3070 8GB	8	FP16	6.0 est.	No	0.012
RTX 4060 Ti 8GB	8	FP16	3.9 est.	No	0.01
RX 7600 8GB	8	FP16	3.9 est.	No	0.014
RTX 4060 8GB	8	FP16	3.6 est.	No	0.012
Mac mini M4 32GB	32	FP16	2.8 est.	Yes	0.003

Fit checks

RTX 5090 32GB — fits · ≈150.0 tok/s Q4 est.
RTX 3090 Ti 24GB — fits · ≈84.4 tok/s Q4 est.
RTX 4090 24GB — fits · ≈84.4 tok/s Q4 est.
RTX 5080 16GB — fits · ≈80.4 tok/s Q4 est.
RTX 6000 Ada 48GB — fits · ≈80.4 tok/s Q4 est.
RX 7900 XTX 24GB — fits · ≈80.4 tok/s Q4 est.
RTX 3090 24GB — fits · ≈78.4 tok/s Q4 est.
RTX 5070 Ti 16GB — fits · ≈75.0 tok/s Q4 est.
L40S 48GB — fits · ≈72.3 tok/s Q4 est.
Mac Studio M3 Ultra 96GB — fits · ≈68.6 tok/s Q4 est.
Mac Studio M2 Ultra 192GB — fits · ≈28.0 tok/s Q4 measured
RX 7900 XT 20GB — fits · ≈67.0 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: Qwen

Qwen3 Qwen3-8B · Qwen3-14B
Qwen3 2507 Qwen3-4B-Instruct-2507 · Qwen3-4B-Thinking-2507 · Qwen3-30B-A3B-Instruct-2507 · Qwen3-235B-A22B-Instruct-2507
Qwen2.5 Qwen2.5-14B-Instruct · Qwen2.5-32B-Instruct · Qwen2.5-72B-Instruct
Qwen2.5 Coder Qwen2.5-Coder-32B-Instruct
Qwen3 Next Qwen3-Next-80B-A3B-Instruct
Qwen3 Coder Qwen3-Coder-480B-A35B-Instruct · Qwen3-Coder-Next
Qwen3.5 Qwen3.5-2B · Qwen3.5-9B · Qwen3.5-27B · Qwen3.5-35B-A3B · Qwen3.5-122B-A10B · Qwen3.5-397B-A17B
Qwen3.6 Qwen3.6-27B · Qwen3.6-35B-A3B

All Qwen models →