Qwen3-235B-A22B-Instruct-2507

Alibaba · 235.1B (active 22B) · MoE · ctx 256K · apache-2.0 · HuggingFace

Qwen3-235B-A22B-Instruct-2507 (Alibaba) is a 235.1B parameters MoE (22B active per token) open-weight model tracked on AI Hashrate. At 8K context we estimate roughly 134.71 GB VRAM for Q4 (weights 129.31 GB + KV 4.4 GB + 1 GB overhead) and 475.6 GB for FP16. About 3 GPUs/accelerators in our catalog fully fit this model at Q4 under that context assumption. Fastest Q4 configs: B200 192GB SXM ≈ 231.4 tok/s (estimated); MI300X 192GB ≈ 153.3 tok/s (estimated); Mac Studio M2 Ultra 192GB ≈ 23.1 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.

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

Hardware	VRAM	Quant	tok/s	Fits?	tok/s/$
H200 141GB SXM5	141	Q4	434.6 est.	No	0.012
Gaudi 3 128GB	128	Q4	244.4 est.	No	0.02
B200 192GB SXM	192	Q4	231.4 est.	Yes	0.006
MI300X 192GB	192	Q4	153.3 est.	Yes	0.01
H100 80GB SXM5	80	Q4	97.6 est.	No	0.003
A100 80GB SXM4	80	Q4	59.4 est.	No	0.004
MacBook Pro M4 Max 128GB	128	Q4	40.7 est.	No	0.009
Mac Studio M3 Ultra 96GB	96	Q4	34.4 est.	No	0.009
MacBook Pro M3 Max 128GB	128	Q4	29.8 est.	No	0.007
B200 192GB SXM	192	FP16	29.6 est.	No	0.001
Mac Studio M2 Ultra 192GB	192	Q4	23.1 est.	Yes	0.003
MI300X 192GB	192	FP16	19.6 est.	No	0.001
RTX 6000 Ada 48GB	48	Q4	10.1 est.	No	0.001
H200 141GB SXM5	141	FP16	9.6 est.	No	0.0
L40S 48GB	48	Q4	9.1 est.	No	0.001
RTX A6000 48GB	48	Q4	8.1 est.	No	0.002
Gaudi 3 128GB	128	FP16	5.4 est.	No	0.0
Mac mini M4 Pro 64GB	64	Q4	5.1 est.	No	0.002
Mac Studio M2 Ultra 192GB	192	FP16	3.0 est.	No	0.0
MacBook Pro M4 Max 128GB	128	FP16	0.9 est.	No	0.0
MacBook Pro M3 Max 128GB	128	FP16	0.7 est.	No	0.0

Fit checks

• RTX 6000 Ada 48GB — does not fit · ≈10.1 tok/s Q4 est.
• L40S 48GB — does not fit · ≈9.1 tok/s Q4 est.
• Mac Studio M3 Ultra 96GB — does not fit · ≈34.4 tok/s Q4 est.
• Mac Studio M2 Ultra 192GB — fits · ≈23.1 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
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 · Qwen3.5-122B-A10B · Qwen3.5-397B-A17B
Qwen3.6 Qwen3.6-27B · Qwen3.6-35B-A3B

All Qwen models →