How much VRAM does Bonsai-8B-unpacked need?

Q4_K_M is exactly 5,198,494,208 bytes (4.84 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is Bonsai-8B-unpacked's KV cache?

4.50 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.

prism-ml · text

Bonsai-8B-unpacked

prism-ml/Bonsai-8B-unpacked

Bonsai-8B-unpacked at Q4_K_M is exactly 5,198,494,208 bytes (4.84 GiB / 5.20 GB) — an effective 5.079 bits per weight, not the nominal 4. Its KV cache at 32K is 4.50 GiB.

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

Parameters

8.2B

Architecture

qwen3

36 layers

Context

65,536

native (config.json)

License

apache-2.0

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
IQ2_M	3.13 GiB	3,363,096,768	3.286	—	bartowski
Q2_K	3.19 GiB	3,429,106,496	3.350	—	bartowski
IQ3_XXS	3.30 GiB	3,547,711,680	3.466	—	bartowski
IQ3_XS	3.52 GiB	3,778,855,264	3.692	—	bartowski
Q3_K_S	3.62 GiB	3,886,203,232	3.797	—	bartowski
IQ3_M	3.76 GiB	4,032,086,368	3.939	—	bartowski
Q2_K_L	3.76 GiB	4,035,782,496	3.943	—	bartowski
Q3_K_M	3.96 GiB	4,249,141,600	4.151	399	bartowski
Q3_K_L	4.17 GiB	4,472,488,288	4.370	—	bartowski
IQ4_XS	4.35 GiB	4,667,703,168	4.560	399	bartowski
Q4_0	4.53 GiB	4,863,670,784	4.752	399	bartowski
Q4_K_S	4.55 GiB	4,885,690,880	4.773	—	bartowski
IQ4_NL	4.55 GiB	4,890,016,256	4.777	—	bartowski
Q4_K_M	4.84 GiB	5,198,494,208	5.079	399	bartowski
Q4_1	4.94 GiB	5,305,151,232	5.183	—	bartowski
Q4_K_L	5.27 GiB	5,659,567,968	5.529	—	bartowski
Q5_K_S	5.36 GiB	5,759,214,592	5.627	—	bartowski
Q5_K_M	5.58 GiB	5,989,835,776	5.852	399	bartowski
Q5_K_L	5.94 GiB	6,373,255,008	6.226	—	bartowski
Q6_K	6.52 GiB	6,998,367,392	6.837	399	bartowski
Q6_K_L	6.80 GiB	7,299,278,688	7.131	—	bartowski
Q8_0	8.11 GiB	8,707,188,576	8.507	399	bartowski
BF16	15.26 GiB	16,383,663,168	16.006	—	bartowski

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.56 GiB	0.56 GiB	—	36 / 0 / 0
8,192	1.13 GiB	1.13 GiB	—	36 / 0 / 0
16,384	2.25 GiB	2.25 GiB	—	36 / 0 / 0
32,768	4.50 GiB	4.50 GiB	—	36 / 0 / 0
65,536	9.00 GiB	9.00 GiB	—	36 / 0 / 0
131,072	18.00 GiB	18.00 GiB	—	36 / 0 / 0

Compare with

same modality, comparable size

vs Qwen3-Coder-30B-A3B-Instruct vs gemma-4-12B-it-qat-q4_0-unquantized vs gemma-4-E4B-it vs Qwen3-30B-A3B-Thinking-2507 vs Qwen3-4B vs Qwen3-8B 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 4.29 GiB. The real file is 4.84 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

128

Hidden size

4096

Vocab

151,669

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

false

Questions people ask

How much VRAM does Bonsai-8B-unpacked need?: Q4_K_M is exactly 5,198,494,208 bytes (4.84 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Bonsai-8B-unpacked's KV cache?: 4.50 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 Bonsai-8B-unpacked 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.