Qwen3.5-122B-A10B

Alibaba · 122.0B (active 10.0B) · MoE · ctx 256K · apache-2.0 · HuggingFace

Qwen3.5-122B-A10B (Alibaba) is a 122.0B parameters MoE (10.0B active per token) open-weight model tracked on AI Hashrate. At 8K context we estimate roughly 70.1 GB VRAM for Q4 (weights 67.1 GB + KV 2.0 GB + 1 GB overhead) and 247.0 GB for FP16. About 10 GPUs/accelerators in our catalog fully fit this model at Q4 under that context assumption. Fastest Q4 configs: B200 192GB SXM ≈ 509.1 tok/s (estimated); MI300X 192GB ≈ 337.3 tok/s (estimated); H200 141GB SXM5 ≈ 305.5 tok/s (estimated); H100 80GB SXM5 ≈ 213.2 tok/s (estimated); Gaudi 3 128GB ≈ 208.5 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.

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

Hardware	VRAM	Quant	tok/s	Fits?	tok/s/$
B200 192GB SXM	192	Q4	509.1 est.	Yes	0.013
MI300X 192GB	192	Q4	337.3 est.	Yes	0.022
H200 141GB SXM5	141	Q4	305.5 est.	Yes	0.009
B200 192GB SXM	192	FP16	241.7 est.	No	0.006
H100 80GB SXM5	80	Q4	213.2 est.	Yes	0.007
Gaudi 3 128GB	128	Q4	208.5 est.	Yes	0.017
MI300X 192GB	192	FP16	160.1 est.	No	0.011
A100 80GB SXM4	80	Q4	129.8 est.	Yes	0.009
RTX 6000 Ada 48GB	48	Q4	81.8 est.	No	0.012
H200 141GB SXM5	141	FP16	78.2 est.	No	0.002
L40S 48GB	48	Q4	73.7 est.	No	0.009
RTX 5090 32GB	32	Q4	67.9 est.	No	0.034
RTX A6000 48GB	48	Q4	65.5 est.	No	0.015
Mac Studio M3 Ultra 96GB	96	Q4	52.1 est.	Yes	0.013
Mac Studio M2 Ultra 192GB	192	Q4	50.9 est.	Yes	0.007
Gaudi 3 128GB	128	FP16	44.0 est.	No	0.004
Mac mini M4 Pro 64GB	64	Q4	41.4 est.	No	0.019
MacBook Pro M4 Max 128GB	128	Q4	34.7 est.	Yes	0.008
MacBook Pro M3 Max 128GB	128	Q4	25.5 est.	Yes	0.006
Mac Studio M2 Ultra 192GB	192	FP16	24.2 est.	No	0.003
RTX 4090 24GB	24	Q4	21.5 est.	No	0.013
RTX 4090D 24GB (China)	24	Q4	21.5 est.	No	0.013
RTX 3090 Ti 24GB	24	Q4	21.5 est.	No	0.011
RX 7900 XTX 24GB	24	Q4	20.5 est.	No	0.021
RTX 3090 24GB	24	Q4	19.9 est.	No	0.013
H100 80GB SXM5	80	FP16	17.6 est.	No	0.001
RX 7900 XT 20GB	20	Q4	11.8 est.	No	0.013
A100 80GB SXM4	80	FP16	10.7 est.	No	0.001
MacBook Pro M4 Max 128GB	128	FP16	7.3 est.	No	0.002
Mac Studio M3 Ultra 96GB	96	FP16	6.2 est.	No	0.002
MacBook Pro M3 Max 128GB	128	FP16	5.4 est.	No	0.001
Mac mini M4 32GB	32	Q4	4.5 est.	No	0.005
Mac mini M4 Pro 64GB	64	FP16	0.9 est.	No	0.0

Fit checks

• RTX 5090 32GB — does not fit · ≈67.9 tok/s Q4 est.
• RTX 3090 Ti 24GB — does not fit · ≈21.5 tok/s Q4 est.
• RTX 4090 24GB — does not fit · ≈21.5 tok/s Q4 est.
• RTX 6000 Ada 48GB — does not fit · ≈81.8 tok/s Q4 est.
• RX 7900 XTX 24GB — does not fit · ≈20.5 tok/s Q4 est.
• RTX 3090 24GB — does not fit · ≈19.9 tok/s Q4 est.
• L40S 48GB — does not fit · ≈73.7 tok/s Q4 est.
• Mac Studio M3 Ultra 96GB — fits · ≈52.1 tok/s Q4 est.
• Mac Studio M2 Ultra 192GB — fits · ≈50.9 tok/s Q4 est.
• RX 7900 XT 20GB — does not fit · ≈11.8 tok/s Q4 est.

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

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 · Qwen3.5-35B-A3B · Qwen3.5-397B-A17B
Qwen3.6 Qwen3.6-27B · Qwen3.6-35B-A3B

All Qwen models →