How much VRAM does SmolVLM-Instruct need?

Q4_K_M is exactly 1,112,242,368 bytes (1.04 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is SmolVLM-Instruct's KV cache?

6.00 GiB at 32K context with an f16 cache, computed per layer from the model's architecture. Quantizing the cache to q8_0 roughly halves it.

HuggingFaceTB · text

SmolVLM-Instruct

HuggingFaceTB/SmolVLM-Instruct

SmolVLM-Instruct at Q4_K_M is exactly 1,112,242,368 bytes (1.04 GiB / 1.11 GB) — an effective 3.961 bits per weight, not the nominal 4. Its KV cache at 32K is 6.00 GiB.

From the file· summed from 1 file(s)From the file· KV per layer

Parameters

2.2B

Architecture

llama

24 layers

Context

16,384

native (config.json)

License

apache-2.0

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
Q4_K_M	1.04 GiB	1,112,242,368	3.961	—	ggml-org
Q8_0	1.80 GiB	1,927,383,680	6.864	—	ggml-org
F16	3.38 GiB	3,626,088,320	12.914	—	ggml-org

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.75 GiB	0.75 GiB	—	24 / 0 / 0
8,192	1.50 GiB	1.50 GiB	—	24 / 0 / 0
16,384	3.00 GiB	3.00 GiB	—	24 / 0 / 0
32,768	6.00 GiB	6.00 GiB	—	24 / 0 / 0
65,536	12.00 GiB	12.00 GiB	—	24 / 0 / 0
131,072	24.00 GiB	24.00 GiB	—	24 / 0 / 0

Compare with

same modality, comparable size

vs Qwen3-4B vs Qwen3-8B vs HyperCLOVAX-SEED-Text-Instruct-1.5B vs Llama-3.2-1B-Instruct vs Qwen3-VL-8B-Instruct-abliterated-v1 vs Llama-3.1-8B-Instruct All alternatives →Will it fit on my card? →

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 1.18 GiB. The real file is 1.04 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from config.json

Layers

Attention heads

KV heads

Head dim

Hidden size

2048

Vocab

49,155

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does SmolVLM-Instruct need?: Q4_K_M is exactly 1,112,242,368 bytes (1.04 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is SmolVLM-Instruct's KV cache?: 6.00 GiB at 32K context with an f16 cache, computed per layer. Quantizing the cache to q8_0 roughly halves it, which is often the difference between a context length fitting and not.
Which quantization of SmolVLM-Instruct should I use?: Q4_K_M is the usual default. Pick the largest quantization that fits your card at the context you actually need — the table above gives exact sizes for every one published.