Qwen3.6-27B

Alibaba · 27.8B · ctx 256K · apache-2.0 · HuggingFace

Qwen3.6-27B (Alibaba) is a 27.8B parameters open-weight model tracked on AI Hashrate. At 8K context we estimate roughly 21.85 GB VRAM for Q4 (weights 15.29 GB + KV 5.56 GB + 1 GB overhead) and 62.16 GB for FP16. About 21 GPUs/accelerators in our catalog fully fit this model at Q4 under that context assumption. Fastest Q4 configs: B200 192GB SXM ≈ 183.1 tok/s (estimated); MI300X 192GB ≈ 121.3 tok/s (estimated); H200 141GB SXM5 ≈ 109.9 tok/s (estimated); H100 80GB SXM5 ≈ 76.7 tok/s (estimated); Gaudi 3 128GB ≈ 75.0 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.

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

Hardware	VRAM	Quant	tok/s	Fits?	tok/s/$
B200 192GB SXM	192	Q4	183.1 est.	Yes	0.005
MI300X 192GB	192	Q4	121.3 est.	Yes	0.008
H200 141GB SXM5	141	Q4	109.9 est.	Yes	0.003
H100 80GB SXM5	80	Q4	76.7 est.	Yes	0.003
Gaudi 3 128GB	128	Q4	75.0 est.	Yes	0.006
B200 192GB SXM	192	FP16	50.4 est.	Yes	0.001
A100 80GB SXM4	80	Q4	46.7 est.	Yes	0.003
RX 7900 XT 20GB	20	Q4	43.8 est.	No	0.049
RTX 5090 32GB	32	Q4	41.0 est.	Yes	0.021
RTX 5080 16GB	16	Q4	33.7 est.	No	0.034
MI300X 192GB	192	FP16	33.4 est.	Yes	0.002
RTX 5070 Ti 16GB	16	Q4	31.4 est.	No	0.042
H200 141GB SXM5	141	FP16	30.2 est.	Yes	0.001
RTX 4080 Super 16GB	16	Q4	25.8 est.	No	0.026
RTX 4070 Ti Super 16GB	16	Q4	23.6 est.	No	0.03
RTX 4090 24GB	24	Q4	23.1 est.	Yes	0.014
RTX 4090D 24GB (China)	24	Q4	23.1 est.	Yes	0.014
RTX 3090 Ti 24GB	24	Q4	23.1 est.	Yes	0.012
RX 9070 16GB	16	Q4	22.4 est.	No	0.041
RX 9070 XT 16GB	16	Q4	22.4 est.	No	0.037
RX 7900 XTX 24GB	24	Q4	22.0 est.	Yes	0.022
RTX 6000 Ada 48GB	48	Q4	22.0 est.	Yes	0.003
RX 7800 XT 16GB	16	Q4	21.9 est.	No	0.044
RTX 3090 24GB	24	Q4	21.4 est.	Yes	0.014
H100 80GB SXM5	80	FP16	21.1 est.	Yes	0.001
Gaudi 3 128GB	128	FP16	20.6 est.	Yes	0.002
RX 7900 GRE 16GB	16	Q4	20.2 est.	No	0.037
L40S 48GB	48	Q4	19.8 est.	Yes	0.002
Arc A770 16GB	16	Q4	19.6 est.	No	0.056
Mac Studio M3 Ultra 96GB	96	Q4	18.7 est.	Yes	0.005
Mac Studio M2 Ultra 192GB	192	Q4	18.3 est.	Yes	0.003
RTX A6000 48GB	48	Q4	17.6 est.	Yes	0.004
RTX 5060 Ti 16GB	16	Q4	15.7 est.	No	0.037
RTX 5070 12GB	12	Q4	13.3 est.	No	0.024
A100 80GB SXM4	80	FP16	12.8 est.	Yes	0.001
MacBook Pro M4 Max 128GB	128	Q4	12.5 est.	Yes	0.003
T4 16GB	16	Q4	11.2 est.	No	0.006
RTX 4070 Ti 12GB	12	Q4	10.9 est.	No	0.014
RTX 3080 10GB	10	Q4	10.4 est.	No	0.015
RTX 6000 Ada 48GB	48	FP16	10.3 est.	No	0.002
RTX 4060 Ti 16GB	16	Q4	10.1 est.	No	0.02
RX 7600 XT 16GB	16	Q4	10.1 est.	No	0.031
RTX 4070 12GB	12	Q4	9.9 est.	No	0.017
RTX 4070 Super 12GB	12	Q4	9.9 est.	No	0.017
L40S 48GB	48	FP16	9.3 est.	No	0.001
MacBook Pro M3 Max 128GB	128	Q4	9.2 est.	Yes	0.002
Arc B580 12GB	12	Q4	9.0 est.	No	0.036
RTX 5090 32GB	32	FP16	8.5 est.	No	0.004
RX 7700 XT 12GB	12	Q4	8.5 est.	No	0.019
RTX A6000 48GB	48	FP16	8.2 est.	No	0.002
RTX 3060 12GB	12	Q4	7.1 est.	No	0.022
Mac mini M4 Pro 64GB	64	Q4	6.2 est.	Yes	0.003
Arc B570 10GB	10	Q4	5.2 est.	No	0.024
Mac mini M4 Pro 64GB	64	FP16	5.2 est.	No	0.002
Mac Studio M3 Ultra 96GB	96	FP16	5.2 est.	Yes	0.001
Mac Studio M2 Ultra 192GB	192	FP16	5.0 est.	Yes	0.001
RTX 5060 8GB	8	Q4	3.9 est.	No	0.013
RTX 5060 Ti 8GB	8	Q4	3.9 est.	No	0.01
RTX 3070 8GB	8	Q4	3.9 est.	No	0.008
MacBook Pro M4 Max 128GB	128	FP16	3.4 est.	Yes	0.001
RTX 4090 24GB	24	FP16	2.7 est.	No	0.002
RTX 4090D 24GB (China)	24	FP16	2.7 est.	No	0.002
RTX 3090 Ti 24GB	24	FP16	2.7 est.	No	0.001
Mac mini M4 32GB	32	Q4	2.7 est.	Yes	0.003
RX 7900 XTX 24GB	24	FP16	2.6 est.	No	0.003
RTX 3090 24GB	24	FP16	2.5 est.	No	0.002
MacBook Pro M3 Max 128GB	128	FP16	2.5 est.	Yes	0.001
RTX 4060 Ti 8GB	8	Q4	2.5 est.	No	0.006
RX 7600 8GB	8	Q4	2.5 est.	No	0.009
RTX 4060 8GB	8	Q4	2.4 est.	No	0.008
RX 7900 XT 20GB	20	FP16	1.5 est.	No	0.002
RTX 5080 16GB	16	FP16	1.1 est.	No	0.001
RTX 5070 Ti 16GB	16	FP16	1.1 est.	No	0.001
RTX 4080 Super 16GB	16	FP16	0.9 est.	No	0.001
RTX 4070 Ti Super 16GB	16	FP16	0.8 est.	No	0.001
RX 9070 16GB	16	FP16	0.8 est.	No	0.001
RX 9070 XT 16GB	16	FP16	0.8 est.	No	0.001
RX 7800 XT 16GB	16	FP16	0.7 est.	No	0.001
RX 7900 GRE 16GB	16	FP16	0.7 est.	No	0.001
Arc A770 16GB	16	FP16	0.7 est.	No	0.002
Mac mini M4 32GB	32	FP16	0.6 est.	No	0.001
RTX 5060 Ti 16GB	16	FP16	0.5 est.	No	0.001
T4 16GB	16	FP16	0.4 est.	No	0.0
RTX 4060 Ti 16GB	16	FP16	0.3 est.	No	0.001
RX 7600 XT 16GB	16	FP16	0.3 est.	No	0.001

Fit checks

• RTX 5090 32GB — fits · ≈41.0 tok/s Q4 est.
• RTX 3090 Ti 24GB — fits · ≈23.1 tok/s Q4 est.
• RTX 4090 24GB — fits · ≈23.1 tok/s Q4 est.
• RTX 5080 16GB — does not fit · ≈33.7 tok/s Q4 est.
• RTX 6000 Ada 48GB — fits · ≈22.0 tok/s Q4 est.
• RX 7900 XTX 24GB — fits · ≈22.0 tok/s Q4 est.
• RTX 3090 24GB — fits · ≈21.4 tok/s Q4 est.
• RTX 5070 Ti 16GB — does not fit · ≈31.4 tok/s Q4 est.
• L40S 48GB — fits · ≈19.8 tok/s Q4 est.
• Mac Studio M3 Ultra 96GB — fits · ≈18.7 tok/s Q4 est.
• Mac Studio M2 Ultra 192GB — fits · ≈18.3 tok/s Q4 est.
• RX 7900 XT 20GB — does not fit · ≈43.8 tok/s Q4 est.

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