Card × model estimate

Phi-4 reasoning mini 3.8B on RTX 3070 Q4_K_M

should fit

On an RTX 3070 (8 GB), Phi-4 reasoning mini 3.8B in Q4_K_M needs about 3,8 Go of video memory: it should fit.

Not measured yet: the figures below are estimates, computed from the models already measured on this card. Measure my card

—

Generation (estimated)

≥ 3,8 Go

Recommended VRAM (8 GB avail.)

—

Power (ref. model)

—

Electricity / M tokens · 0,25 €/kWh

No model has been measured on the RTX 3070 yet: no speed estimate possible.

What about a longer context?

Context	KV cache	Estimated VRAM	Fits in 8 GB?	Estimated speed
4 096	0,5 Go	3,3 Go	yes	—
8 192	1,0 Go	3,8 Go	yes	—
16 384	2,0 Go	4,8 Go	yes	—
32 768	4,0 Go	6,8 Go	yes	—
65 536	8,0 Go	10,8 Go	no	—
131 072	16,0 Go	18,8 Go	no	—

Estimated VRAM = file + KV cache (full-attention layers, f16) + ~0,5 Go of buffers, excluding memory used by the system — hence the slightly higher “recommended VRAM”, which keeps that margin. Speed ≈ estimated speed × weights read ÷ (weights read + added KV cache).

Phi-4 reasoning mini 3.8B on other cards

should fit

Phi-4 reasoning mini 3.8B page RTX 3070 page Feel the speed Download the GGUF (2,3 Go)

Estimates based on real measurements of other models on this card and on the model's public data (Microsoft Phi). As soon as a measurement exists, this page shows the real figures.