Can T4 16GB run Phi-4-mini-instruct?

NVIDIA · 16 GB GDDR6 · 320 GB/s bandwidth · Microsoft · 3.8B · model ctx up to 128K

Yes — the T4 16GB (16 GB GDDR6) can run Phi-4-mini-instruct. At Q4 with the default 8K context it needs ≈ 3.85 GB VRAM (weights 2.09 GB + KV cache 0.76 GB + 1 GB runtime overhead), which fits within 16 GB. Decode at Q4/8K: ≈ 53.6 tok/s (estimated, batch 1). FP16 (≈ 9.36 GB) also fits, at ≈ 14.7 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	3.47 GB	Yes	53.6 est.
Q4	8K default	3.85 GB	Yes	53.6 est.
Q4	32K	6.13 GB	Yes	53.6 est.
FP16	4K	8.98 GB	Yes	14.7 est.
FP16	8K default	9.36 GB	Yes	14.7 est.
FP16	32K	11.64 GB	Yes	14.7 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	T4 16GB VRAM
Q4	2.09 GB	0.76 GB	1 GB	3.85 GB	16 GB
FP16	7.6 GB	0.76 GB	1 GB	9.36 GB	16 GB

Other GPUs that run Phi-4-mini-instruct

• B200 192GB SXM — ≈1339.7 tok/s Q4 est.
• MI300X 192GB — ≈887.6 tok/s Q4 est.
• H200 141GB SXM5 — ≈803.8 tok/s Q4 est.
• H100 80GB SXM5 — ≈561.0 tok/s Q4 est.
• Gaudi 3 128GB — ≈548.6 tok/s Q4 est.
• A100 80GB SXM4 — ≈341.5 tok/s Q4 est.
• RTX 5090 32GB — ≈300.1 tok/s Q4 est.
• RTX 4090 24GB — ≈168.8 tok/s Q4 est.

All GPUs for Phi-4-mini-instruct →

Other models for the T4 16GB

• Qwen3-0.6B — ≈254.5 tok/s Q4 est.
• gemma-3-1b-it — ≈203.6 tok/s Q4 est.
• Llama-3.2-1B-Instruct — ≈169.7 tok/s Q4 est.
• Qwen3-1.7B — ≈101.8 tok/s Q4 est.
• Llama-3.2-3B-Instruct — ≈63.6 tok/s Q4 est.
• gpt-oss-20b — ≈56.6 tok/s Q4 est.
• Phi-4-mini-reasoning — ≈53.6 tok/s Q4 est.
• gemma-3-4b-it — ≈52.2 tok/s Q4 est.

All models on the T4 16GB →