Qwen3.5-35B-A3B

Alibaba · 34.7B (active 3.0B) · MoE · ctx 256K · apache-2.0 · HuggingFace

Qwen3.5-35B-A3B (Alibaba) is a 34.7B parameters MoE (3.0B active per token) open-weight model tracked on AI Hashrate. At 8K context we estimate roughly 20.69 GB VRAM for Q4 (weights 19.09 GB + KV 0.6 GB + 1 GB overhead) and 71.0 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 ≈ 1697.0 tok/s (estimated); MI300X 192GB ≈ 1124.2 tok/s (estimated); H200 141GB SXM5 ≈ 1018.2 tok/s (estimated); H100 80GB SXM5 ≈ 710.6 tok/s (estimated); Gaudi 3 128GB ≈ 694.9 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.

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

Hardware	VRAM	Quant	tok/s	Fits?	tok/s/$
B200 192GB SXM	192	Q4	1697.0 est.	Yes	0.042
MI300X 192GB	192	Q4	1124.2 est.	Yes	0.075
H200 141GB SXM5	141	Q4	1018.2 est.	Yes	0.029
H100 80GB SXM5	80	Q4	710.6 est.	Yes	0.024
Gaudi 3 128GB	128	Q4	694.9 est.	Yes	0.056
B200 192GB SXM	192	FP16	466.7 est.	Yes	0.012
RX 7900 XT 20GB	20	Q4	453.0 est.	No	0.504
A100 80GB SXM4	80	Q4	432.5 est.	Yes	0.029
RTX 5090 32GB	32	Q4	380.1 est.	Yes	0.19
RTX 5080 16GB	16	Q4	347.9 est.	No	0.348
RTX 5070 Ti 16GB	16	Q4	324.7 est.	No	0.434
MI300X 192GB	192	FP16	309.2 est.	Yes	0.021
H200 141GB SXM5	141	FP16	280.0 est.	Yes	0.008
RTX 4080 Super 16GB	16	Q4	266.8 est.	No	0.267
RTX 4070 Ti Super 16GB	16	Q4	243.6 est.	No	0.305
RX 9070 16GB	16	Q4	232.0 est.	No	0.423
RX 9070 XT 16GB	16	Q4	232.0 est.	No	0.387
RX 7800 XT 16GB	16	Q4	226.2 est.	No	0.453
RTX 4090 24GB	24	Q4	213.8 est.	Yes	0.134
RTX 4090D 24GB (China)	24	Q4	213.8 est.	Yes	0.134
RTX 3090 Ti 24GB	24	Q4	213.8 est.	Yes	0.107
RX 7900 GRE 16GB	16	Q4	208.8 est.	No	0.38
RX 7900 XTX 24GB	24	Q4	203.6 est.	Yes	0.204
RTX 6000 Ada 48GB	48	Q4	203.6 est.	Yes	0.03
Arc A770 16GB	16	Q4	203.0 est.	No	0.582
RTX 3090 24GB	24	Q4	198.5 est.	Yes	0.132
H100 80GB SXM5	80	FP16	195.4 est.	Yes	0.007
Gaudi 3 128GB	128	FP16	191.1 est.	Yes	0.015
L40S 48GB	48	Q4	183.3 est.	Yes	0.023
Mac Studio M3 Ultra 96GB	96	Q4	173.7 est.	Yes	0.043
Mac Studio M2 Ultra 192GB	192	Q4	169.7 est.	Yes	0.024
RTX A6000 48GB	48	Q4	162.9 est.	Yes	0.036
RTX 5060 Ti 16GB	16	Q4	162.4 est.	No	0.379
RTX 5070 12GB	12	Q4	137.0 est.	No	0.25
A100 80GB SXM4	80	FP16	118.9 est.	Yes	0.008
T4 16GB	16	Q4	116.0 est.	No	0.058
MacBook Pro M4 Max 128GB	128	Q4	115.8 est.	Yes	0.026
RTX 4070 Ti 12GB	12	Q4	112.9 est.	No	0.141
RTX 3080 10GB	10	Q4	107.6 est.	No	0.154
RTX 4060 Ti 16GB	16	Q4	104.4 est.	No	0.209
RX 7600 XT 16GB	16	Q4	104.4 est.	No	0.317
RTX 4070 12GB	12	Q4	102.8 est.	No	0.172
RTX 4070 Super 12GB	12	Q4	102.8 est.	No	0.172
Arc B580 12GB	12	Q4	93.0 est.	No	0.373
RX 7700 XT 12GB	12	Q4	88.1 est.	No	0.196
MacBook Pro M3 Max 128GB	128	Q4	84.8 est.	Yes	0.02
RTX 3060 12GB	12	Q4	73.4 est.	No	0.223
RTX 6000 Ada 48GB	48	FP16	73.1 est.	No	0.011
L40S 48GB	48	FP16	65.8 est.	No	0.008
RTX 5090 32GB	32	FP16	60.7 est.	No	0.03
RTX A6000 48GB	48	FP16	58.5 est.	No	0.013
Mac mini M4 Pro 64GB	64	Q4	57.9 est.	Yes	0.026
Arc B570 10GB	10	Q4	53.8 est.	No	0.246
Mac Studio M3 Ultra 96GB	96	FP16	47.8 est.	Yes	0.012
Mac Studio M2 Ultra 192GB	192	FP16	46.7 est.	Yes	0.007
RTX 5060 8GB	8	Q4	40.6 est.	No	0.136
RTX 5060 Ti 8GB	8	Q4	40.6 est.	No	0.107
RTX 3070 8GB	8	Q4	40.6 est.	No	0.081
Mac mini M4 Pro 64GB	64	FP16	37.0 est.	No	0.017
MacBook Pro M4 Max 128GB	128	FP16	31.8 est.	Yes	0.007
RTX 4060 Ti 8GB	8	Q4	26.1 est.	No	0.065
RX 7600 8GB	8	Q4	26.1 est.	No	0.097
Mac mini M4 32GB	32	Q4	25.5 est.	Yes	0.026
RTX 4060 8GB	8	Q4	24.6 est.	No	0.082
MacBook Pro M3 Max 128GB	128	FP16	23.3 est.	Yes	0.006
RTX 4090 24GB	24	FP16	19.2 est.	No	0.012
RTX 4090D 24GB (China)	24	FP16	19.2 est.	No	0.012
RTX 3090 Ti 24GB	24	FP16	19.2 est.	No	0.01
RX 7900 XTX 24GB	24	FP16	18.3 est.	No	0.018
RTX 3090 24GB	24	FP16	17.8 est.	No	0.012
RX 7900 XT 20GB	20	FP16	10.6 est.	No	0.012
Mac mini M4 32GB	32	FP16	4.1 est.	No	0.004

Fit checks

• RTX 5090 32GB — fits · ≈380.1 tok/s Q4 est.
• RTX 3090 Ti 24GB — fits · ≈213.8 tok/s Q4 est.
• RTX 4090 24GB — fits · ≈213.8 tok/s Q4 est.
• RTX 5080 16GB — does not fit · ≈347.9 tok/s Q4 est.
• RTX 6000 Ada 48GB — fits · ≈203.6 tok/s Q4 est.
• RX 7900 XTX 24GB — fits · ≈203.6 tok/s Q4 est.
• RTX 3090 24GB — fits · ≈198.5 tok/s Q4 est.
• RTX 5070 Ti 16GB — does not fit · ≈324.7 tok/s Q4 est.
• L40S 48GB — fits · ≈183.3 tok/s Q4 est.
• Mac Studio M3 Ultra 96GB — fits · ≈173.7 tok/s Q4 est.
• Mac Studio M2 Ultra 192GB — fits · ≈169.7 tok/s Q4 est.
• RX 7900 XT 20GB — does not fit · ≈453.0 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 Qwen2.5-7B-Instruct · 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 · Qwen3.5-122B-A10B · Qwen3.5-397B-A17B
Qwen3.6 Qwen3.6-27B · Qwen3.6-35B-A3B

All Qwen models →