How much VRAM does TinyLlama-1.1B-Chat-v0.3 need?

Q4_K_M is exactly 667,822,976 bytes (0.62 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is TinyLlama-1.1B-Chat-v0.3's KV cache?

0.69 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.

TinyLlama · text

TinyLlama-1.1B-Chat-v0.3

TinyLlama/TinyLlama-1.1B-Chat-v0.3

TinyLlama-1.1B-Chat-v0.3 at Q4_K_M is exactly 667,822,976 bytes (0.62 GiB / 0.67 GB) — an effective 4.857 bits per weight, not the nominal 4. Its KV cache at 32K is 0.69 GiB.

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

Parameters

1.1B

Architecture

llama

22 layers

Context

2,048

native (config.json)

License

apache-2.0

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
Q2_K	0.45 GiB	482,149,856	3.506	—	TheBloke
Q3_K_S	0.47 GiB	499,349,600	3.631	—	TheBloke
Q3_K_M	0.51 GiB	549,853,280	3.999	201	TheBloke
Q3_K_L	0.55 GiB	591,534,176	4.302	—	TheBloke
Q4_0	0.59 GiB	636,734,336	4.630	201	TheBloke
Q4_K_S	0.60 GiB	642,763,648	4.674	—	TheBloke
Q4_K_M	0.62 GiB	667,822,976	4.857	201	TheBloke
Q5_0	0.71 GiB	766,037,632	5.571	—	TheBloke
Q5_K_S	0.71 GiB	766,037,632	5.571	—	TheBloke
Q5_K_M	0.73 GiB	782,052,992	5.687	201	TheBloke
Q6_K	0.84 GiB	903,422,400	6.570	201	TheBloke
Q8_0	1.09 GiB	1,169,820,992	8.507	201	TheBloke

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.09 GiB	0.09 GiB	—	22 / 0 / 0
8,192	0.17 GiB	0.17 GiB	—	22 / 0 / 0
16,384	0.34 GiB	0.34 GiB	—	22 / 0 / 0
32,768	0.69 GiB	0.69 GiB	—	22 / 0 / 0
65,536	1.38 GiB	1.38 GiB	—	22 / 0 / 0
131,072	2.75 GiB	2.75 GiB	—	22 / 0 / 0

Compare with

same modality, comparable size

vs Qwen3-4B vs HyperCLOVAX-SEED-Text-Instruct-1.5B vs Llama-3.2-1B-Instruct vs ced-base vs gemma-3-1b-it vs Qwen3-1.7B 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 0.58 GiB. The real file is 0.62 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

32,003

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does TinyLlama-1.1B-Chat-v0.3 need?: Q4_K_M is exactly 667,822,976 bytes (0.62 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is TinyLlama-1.1B-Chat-v0.3's KV cache?: 0.69 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 TinyLlama-1.1B-Chat-v0.3 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.