How much VRAM does orpheus-3b-0.1-ft need?

Q4_K_M is exactly 2,092,567,712 bytes (1.95 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is orpheus-3b-0.1-ft's KV cache?

3.50 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.

canopylabs · audio tts

orpheus-3b-0.1-ft

canopylabs/orpheus-3b-0.1-ft

orpheus-3b-0.1-ft at Q4_K_M is exactly 2,092,567,712 bytes (1.95 GiB / 2.09 GB) — an effective 4.425 bits per weight, not the nominal 4. Its KV cache at 32K is 3.50 GiB.

From the file· summed from 1 file(s)From the file· KV from mirror (mirror:unsloth/orpheus-3b-0.1-ft)

Parameters

3.8B

Architecture

llama

28 layers

Context

131,072

native (config.json)

License

apache-2.0

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
UD-IQ1_S	0.91 GiB	973,833,984	2.059	—	unsloth
UD-IQ1_M	0.95 GiB	1,021,904,640	2.161	—	unsloth
UD-IQ2_XXS	1.03 GiB	1,108,068,096	2.343	—	unsloth
UD-IQ2_M	1.23 GiB	1,317,750,528	2.787	—	unsloth
UD-IQ3_XXS	1.33 GiB	1,431,635,712	3.027	—	unsloth
Q2_K	1.34 GiB	1,437,127,200	3.039	—	unsloth
Q2_K_L	1.34 GiB	1,437,127,200	3.039	—	unsloth
Q3_K_S	1.51 GiB	1,616,040,480	3.417	—	unsloth
Q3_K_M	1.64 GiB	1,760,350,752	3.723	255	unsloth
IQ4_XS	1.77 GiB	1,902,301,728	4.023	—	unsloth
IQ4_NL	1.85 GiB	1,990,382,112	4.209	—	unsloth
Q4_0	1.86 GiB	1,995,100,704	4.219	255	unsloth
Q4_K_S	1.86 GiB	2,001,392,160	4.232	—	unsloth
Q4_K_M	1.95 GiB	2,092,567,712	4.425	—	vinhnx90
Q4_K_M	1.95 GiB	2,092,569,120	4.425	255	unsloth
Q4_1	2.02 GiB	2,166,542,880	4.582	—	unsloth
Q5_K_S	2.18 GiB	2,342,703,648	4.954	—	unsloth
Q4_K_M	2.20 GiB	2,363,760,768	4.999	—	isaiahbjork
Q5_K_M	2.23 GiB	2,395,345,440	5.066	255	unsloth
Q6_K	2.53 GiB	2,717,045,280	5.746	255	unsloth
Q8_0	3.27 GiB	3,516,431,616	7.436	255	unsloth
F16	6.16 GiB	6,610,830,080	13.980	255	unsloth

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.44 GiB	0.44 GiB	—	28 / 0 / 0
8,192	0.88 GiB	0.88 GiB	—	28 / 0 / 0
16,384	1.75 GiB	1.75 GiB	—	28 / 0 / 0
32,768	3.50 GiB	3.50 GiB	—	28 / 0 / 0
65,536	7.00 GiB	7.00 GiB	—	28 / 0 / 0
131,072	14.00 GiB	14.00 GiB	—	28 / 0 / 0

Compare with

same modality, comparable size

vs Ace-Step1.5 vs Qwen3-TTS-12Hz-0.6B-Base vs OmniVoice vs neutts-air vs Fun-CosyVoice3-0.5B-2512 vs VoxCPM2 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 1.98 GiB. The real file is 1.95 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from mirror:unsloth/orpheus-3b-0.1-ft

Layers

Attention heads

KV heads

Head dim

128

Hidden size

3072

Vocab

156,940

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does orpheus-3b-0.1-ft need?: Q4_K_M is exactly 2,092,567,712 bytes (1.95 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is orpheus-3b-0.1-ft's KV cache?: 3.50 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 orpheus-3b-0.1-ft 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.