Can MI300X 192GB run Qwen2.5-72B-Instruct?

AMD · 192 GB HBM3 · 5300 GB/s bandwidth · Alibaba · 72.7B · model ctx up to 128K

Yes — the MI300X 192GB (192 GB HBM3) can run Qwen2.5-72B-Instruct. At Q4 with the default 8K context it needs ≈ 55.53 GB VRAM (weights 39.99 GB + KV cache 14.54 GB + 1 GB runtime overhead), which fits within 192 GB. Decode at Q4/8K: ≈ 46.4 tok/s (estimated, batch 1). FP16 (≈ 160.94 GB) also fits, at ≈ 12.8 tok/s (estimated). Estimates come from memory-bandwidth math; rows tagged measured override estimates. Relative ranking is more reliable than absolute tok/s. Methodology.

Fit & speed by quant and context

Quant	Context	VRAM needed	Fits?	tok/s (decode)
Q4	4K	48.26 GB	Yes	46.4 est.
Q4	8K default	55.53 GB	Yes	46.4 est.
Q4	32K	99.15 GB	Yes	46.4 est.
FP16	4K	153.67 GB	Yes	12.8 est.
FP16	8K default	160.94 GB	Yes	12.8 est.
FP16	32K	204.56 GB	No	32.1 est.

Fits = weights + KV(ctx) + 1 GB overhead ≤ 95% of VRAM. Measured anchors are context-agnostic; the fit verdict is recomputed per context. A missing tok/s means the model is far beyond this card (offload-only territory).

VRAM breakdown at 8K context

Quant	Weights	KV cache	Overhead	Total needed	MI300X 192GB VRAM
Q4	39.99 GB	14.54 GB	1 GB	55.53 GB	192 GB
FP16	145.4 GB	14.54 GB	1 GB	160.94 GB	192 GB

Other GPUs that run Qwen2.5-72B-Instruct

• B200 192GB SXM — ≈70.0 tok/s Q4 est.
• H200 141GB SXM5 — ≈42.0 tok/s Q4 est.
• H100 80GB SXM5 — ≈29.3 tok/s Q4 est.
• Gaudi 3 128GB — ≈28.7 tok/s Q4 est.
• A100 80GB SXM4 — ≈16.47 tok/s Q4 measured
• MacBook Pro M4 Max 128GB — ≈11.0 tok/s Q4 measured
• Mac Studio M3 Ultra 96GB — ≈7.2 tok/s Q4 est.
• Mac Studio M2 Ultra 192GB — ≈7.0 tok/s Q4 est.

All GPUs for Qwen2.5-72B-Instruct →

Other models for the MI300X 192GB

• Qwen3-0.6B — ≈4215.9 tok/s Q4 est.
• gemma-3-1b-it — ≈3372.7 tok/s Q4 est.
• Llama-3.2-1B-Instruct — ≈2810.6 tok/s Q4 est.
• Qwen3-1.7B — ≈1686.4 tok/s Q4 est.
• Qwen3-30B-A3B — ≈1124.2 tok/s Q4 est.
• Qwen3-Coder-30B-A3B-Instruct — ≈1124.2 tok/s Q4 est.
• GLM-4.7-Flash — ≈1124.2 tok/s Q4 est.
• Qwen3-30B-A3B-Instruct-2507 — ≈1124.2 tok/s Q4 est.

All models on the MI300X 192GB →