Qwen3-Next-80B-A3B-Instruct

Alibaba · 81.3B (active 3B) · MoE · ctx 256K · apache-2.0 · HuggingFace

Qwen3-Next-80B-A3B-Instruct (Alibaba) is a 81.3B parameters MoE (3B active per token) open-weight model tracked on AI Hashrate. At 8K context we estimate roughly 46.32 GB VRAM for Q4 (weights 44.72 GB + KV 0.6 GB + 1 GB overhead) and 164.2 GB for FP16. About 11 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.

56 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
RTX 6000 Ada 48GB	48	Q4	624.8 est.	No	0.092
H200 141GB SXM5	141	FP16	589.9 est.	No	0.017
L40S 48GB	48	Q4	562.3 est.	No	0.07
RTX 5090 32GB	32	Q4	518.3 est.	No	0.259
RTX A6000 48GB	48	Q4	499.8 est.	No	0.111
B200 192GB SXM	192	FP16	466.7 est.	Yes	0.012
A100 80GB SXM4	80	Q4	432.5 est.	Yes	0.029
Gaudi 3 128GB	128	FP16	331.8 est.	No	0.027
MI300X 192GB	192	FP16	309.2 est.	Yes	0.021
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 4090 24GB	24	Q4	164.0 est.	No	0.103
RTX 4090D 24GB (China)	24	Q4	164.0 est.	No	0.103
RTX 3090 Ti 24GB	24	Q4	164.0 est.	No	0.082
RX 7900 XTX 24GB	24	Q4	156.2 est.	No	0.156
RTX 3090 24GB	24	Q4	152.3 est.	No	0.102
H100 80GB SXM5	80	FP16	132.5 est.	No	0.004
MacBook Pro M4 Max 128GB	128	Q4	115.8 est.	Yes	0.026
RX 7900 XT 20GB	20	Q4	90.4 est.	No	0.101
MacBook Pro M3 Max 128GB	128	Q4	84.8 est.	Yes	0.02
A100 80GB SXM4	80	FP16	80.7 est.	No	0.005
RTX 5080 16GB	16	Q4	69.4 est.	No	0.069
RTX 5070 Ti 16GB	16	Q4	64.8 est.	No	0.087
Mac mini M4 Pro 64GB	64	Q4	57.9 est.	Yes	0.026
MacBook Pro M4 Max 128GB	128	FP16	55.3 est.	No	0.013
RTX 4080 Super 16GB	16	Q4	53.2 est.	No	0.053
RTX 4070 Ti Super 16GB	16	Q4	48.6 est.	No	0.061
Mac Studio M2 Ultra 192GB	192	FP16	46.7 est.	Yes	0.007
Mac Studio M3 Ultra 96GB	96	FP16	46.7 est.	No	0.012
RX 9070 16GB	16	Q4	46.3 est.	No	0.084
RX 9070 XT 16GB	16	Q4	46.3 est.	No	0.077
RX 7800 XT 16GB	16	Q4	45.1 est.	No	0.09
RX 7900 GRE 16GB	16	Q4	41.7 est.	No	0.076
MacBook Pro M3 Max 128GB	128	FP16	40.5 est.	No	0.01
Arc A770 16GB	16	Q4	40.5 est.	No	0.116
Mac mini M4 32GB	32	Q4	34.7 est.	No	0.035
RTX 5060 Ti 16GB	16	Q4	32.4 est.	No	0.076
RTX 5070 12GB	12	Q4	27.3 est.	No	0.05
T4 16GB	16	Q4	23.1 est.	No	0.012
RTX 4070 Ti 12GB	12	Q4	22.5 est.	No	0.028
RTX 4060 Ti 16GB	16	Q4	20.8 est.	No	0.042
RX 7600 XT 16GB	16	Q4	20.8 est.	No	0.063
RTX 4070 12GB	12	Q4	20.5 est.	No	0.034
RTX 4070 Super 12GB	12	Q4	20.5 est.	No	0.034
Arc B580 12GB	12	Q4	18.5 est.	No	0.074
RX 7700 XT 12GB	12	Q4	17.6 est.	No	0.039
RTX 3060 12GB	12	Q4	14.6 est.	No	0.044
RTX 6000 Ada 48GB	48	FP16	13.7 est.	No	0.002
L40S 48GB	48	FP16	12.3 est.	No	0.002
RTX A6000 48GB	48	FP16	10.9 est.	No	0.002
Mac mini M4 Pro 64GB	64	FP16	6.9 est.	No	0.003

Fit checks

• RTX 5090 32GB — does not fit · ≈518.3 tok/s Q4 est.
• RTX 3090 Ti 24GB — does not fit · ≈164.0 tok/s Q4 est.
• RTX 4090 24GB — does not fit · ≈164.0 tok/s Q4 est.
• RTX 5080 16GB — does not fit · ≈69.4 tok/s Q4 est.
• RTX 6000 Ada 48GB — does not fit · ≈624.8 tok/s Q4 est.
• RX 7900 XTX 24GB — does not fit · ≈156.2 tok/s Q4 est.
• RTX 3090 24GB — does not fit · ≈152.3 tok/s Q4 est.
• RTX 5070 Ti 16GB — does not fit · ≈64.8 tok/s Q4 est.
• L40S 48GB — does not fit · ≈562.3 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 · ≈90.4 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 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 →