ibm-granite · text

granite-embedding-107m-multilingual

ibm-granite/granite-embedding-107m-multilingual

granite-embedding-107m-multilingual at Q4_K_M is exactly 117,011,136 bytes (0.11 GiB / 0.12 GB) — an effective 8.749 bits per weight, not the nominal 4.

From the file· summed from 1 file(s)From the file· KV per layer
Parameters
107M
Architecture
bert
6 layers
Context
514
native (config.json)
License
apache-2.0

Shipped quantizations

● exact bytes, summed from published files
QuantSize●Exact bytes●Effective bpw●Tensors●Publisher
IQ4_XS0.11 GiB114,785,9208.583—cstr
Q4_K0.11 GiB114,896,5128.591—cstr
IQ3_M0.11 GiB115,462,8488.633—bartowski
IQ4_XS0.11 GiB115,601,0888.643—bartowski
Q4_00.11 GiB115,711,6808.652—bartowski
IQ4_NL0.11 GiB115,711,6808.652—bartowski
Q3_K_M0.11 GiB115,951,2968.670—bartowski
Q3_K_L0.11 GiB116,485,8248.710—bartowski
Q4_K_S0.11 GiB116,633,2808.721—bartowski
Q4_K_M0.11 GiB117,011,1368.749—bartowski
Q4_K_L0.11 GiB117,011,1368.749—bartowski
Q5_K_S0.11 GiB117,481,1528.784—bartowski
Q5_K_M0.11 GiB117,729,9848.803—bartowski
Q5_K_L0.11 GiB117,729,9848.803—bartowski
Q6_K_L0.11 GiB120,163,0088.985—bartowski
Q6_K0.11 GiB120,163,0088.985—bartowski
Q8_00.11 GiB120,204,9288.988—cstr
Q8_00.11 GiB121,020,0969.049—bartowski
F160.21 GiB220,974,08016.522—bartowski

No KV cache

architectural, not a gap in our data

This architecture allocates no KV cache. Encoder and embedding models process their input in one pass rather than generating token by token, so there is nothing to carry forward between steps and memory does not grow with context. Its footprint is the weights plus a working buffer, and that is the whole story.

Compare with

same modality, comparable size
vs ced-basevs embeddinggemma-300mvs Qwen3-0.6Bvs SmolLM2-135M-Instructvs gemma-3-270m-itvs tinygemma3_cifarAll alternatives →Will it fit on my card? →

Will it run on your card?

full quant x context sweep
By VRAM:8GB12GB16GB24GB32GB48GB
Radeon RX 6500 XT 4GBGeForce RTX 3050 6GBGeForce RTX 5050 8GBGeForce RTX 3080 10GBGeForce RTX 2080 Ti 11GBGeForce RTX 5070 12GBGeForce RTX 5060 Ti 16GBApple M3 Pro 18GBGeForce RTX 3080 Ti 20GBGeForce RTX 5090 D V2 24GBGeForce RTX 5090 D 32GBApple M5 Max 36GBApple M5 Pro 48GBApple M5 Max 64GBApple M3 Ultra 96GBApple M5 Max 128GBApple M2 Ultra 192GBApple M3 Ultra 256GBApple M3 Ultra 512GB

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 0.06 GiB. The real file is 0.11 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from config.json
Layers
6
Attention heads
12
KV heads
12
Head dim
32
Hidden size
384
Vocab
250,002
Sliding window
none
SWA period
—
MLA
no
Experts
—
Experts per token
—
use_sliding_window
—

Questions people ask

How much VRAM does granite-embedding-107m-multilingual need?
Q4_K_M is exactly 117,011,136 bytes (0.11 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
Which quantization of granite-embedding-107m-multilingual 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.