Can L40S 48GB run Phi-4?

NVIDIA · 48 GB GDDR6 · 864 GB/s bandwidth · Microsoft · 14.7B · model ctx up to 16K

Yes — the L40S 48GB (48 GB GDDR6) can run Phi-4. At Q4 with the default 8K context it needs ≈ 12.03 GB VRAM (weights 8.09 GB + KV cache 2.94 GB + 1 GB runtime overhead), which fits within 48 GB. Decode at Q4/8K: ≈ 37.4 tok/s (estimated, batch 1). FP16 (≈ 33.34 GB) also fits, at ≈ 10.3 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	10.56 GB	Yes	37.4 est.
Q4	8K default	12.03 GB	Yes	37.4 est.
Q4	32K	20.84 GB	Yes	37.4 est.
FP16	4K	31.87 GB	Yes	10.3 est.
FP16	8K default	33.34 GB	Yes	10.3 est.
FP16	32K	42.16 GB	Yes	10.3 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	L40S 48GB VRAM
Q4	8.09 GB	2.94 GB	1 GB	12.03 GB	48 GB
FP16	29.4 GB	2.94 GB	1 GB	33.34 GB	48 GB

Other GPUs that run Phi-4

• B200 192GB SXM — ≈346.3 tok/s Q4 est.
• MI300X 192GB — ≈229.4 tok/s Q4 est.
• H200 141GB SXM5 — ≈207.8 tok/s Q4 est.
• H100 80GB SXM5 — ≈145.0 tok/s Q4 est.
• Gaudi 3 128GB — ≈141.8 tok/s Q4 est.
• A100 80GB SXM4 — ≈88.3 tok/s Q4 est.
• RTX 5090 32GB — ≈77.6 tok/s Q4 est.
• RTX 4090 24GB — ≈43.6 tok/s Q4 est.

All GPUs for Phi-4 →

Other models for the L40S 48GB

• Qwen3-0.6B — ≈687.3 tok/s Q4 est.
• gemma-3-1b-it — ≈549.8 tok/s Q4 est.
• Llama-3.2-1B-Instruct — ≈458.2 tok/s Q4 est.
• Qwen3-1.7B — ≈274.9 tok/s Q4 est.
• Qwen3-30B-A3B — ≈183.3 tok/s Q4 est.
• Qwen3-Coder-30B-A3B-Instruct — ≈183.3 tok/s Q4 est.
• GLM-4.7-Flash — ≈183.3 tok/s Q4 est.
• Qwen3-30B-A3B-Instruct-2507 — ≈183.3 tok/s Q4 est.

All models on the L40S 48GB →