ibm-granite · text

granite-34b-code-base-8k

ibm-granite/granite-34b-code-base-8k

granite-34b-code-base-8k at Q4_K_M is exactly 21,383,628,160 bytes (19.92 GiB / 21.38 GB) — an effective 5.074 bits per weight, not the nominal 4.

From the file· summed from 1 file(s)From the file· KV per layer
Parameters
33.7B
Architecture
starcoder
88 layers
Context
—
native (config.json)
License
apache-2.0

Shipped quantizations

● exact bytes, summed from published files
QuantSize●Exact bytes●Effective bpw●Tensors●Publisher
IQ1_S6.87 GiB7,377,962,6561.751—bartowski
I1-IQ1_S6.87 GiB7,377,963,1361.751—mradermacher
IQ1_M7.49 GiB8,042,989,2161.908—bartowski
I1-IQ1_M7.49 GiB8,042,989,6961.908—mradermacher
IQ2_XXS8.52 GiB9,151,366,8162.171—bartowski
I1-IQ2_XXS8.52 GiB9,151,367,2962.171—mradermacher
IQ2_XS9.45 GiB10,141,877,9202.406—bartowski
I1-IQ2_XS9.45 GiB10,141,878,4002.406—mradermacher
IQ2_S10.04 GiB10,777,708,1922.557—bartowski
I1-IQ2_S10.04 GiB10,777,708,6722.557—mradermacher
IQ2_M10.86 GiB11,664,410,2722.768—bartowski
I1-IQ2_M10.86 GiB11,664,410,7522.768—mradermacher
Q2_K12.21 GiB13,107,021,4723.110—bartowski
I1-Q2_K12.21 GiB13,107,021,9523.110—mradermacher
IQ3_XXS12.44 GiB13,359,957,6643.170—bartowski
I1-IQ3_XXS12.44 GiB13,359,958,1443.170—mradermacher
IQ3_XS13.36 GiB14,340,834,9763.403—bartowski
I1-IQ3_XS13.36 GiB14,340,835,4563.403—mradermacher
IQ3_S13.79 GiB14,807,975,5843.514—bartowski
Q3_K_S13.79 GiB14,807,975,5843.514—bartowski
I1-IQ3_S13.79 GiB14,807,976,0643.514—mradermacher
I1-Q3_K_S13.79 GiB14,807,976,0643.514—mradermacher
IQ3_M14.84 GiB15,929,329,3123.780—bartowski
I1-IQ3_M14.84 GiB15,929,329,7923.780—mradermacher
Q3_K_M16.36 GiB17,567,860,3844.168—bartowski
I1-Q3_K_M16.36 GiB17,567,860,8644.168—mradermacher
IQ4_XS16.95 GiB18,195,826,3364.317—bartowski
I1-IQ4_XS16.95 GiB18,195,826,8164.317—mradermacher
IQ4_NL17.92 GiB19,238,241,9524.565—bartowski
I1-Q4_018.01 GiB19,342,051,4564.590—mradermacher
Q4_K_S18.11 GiB19,445,860,0004.614—bartowski
I1-Q4_K_S18.11 GiB19,445,860,4804.614—mradermacher
Q3_K_L18.21 GiB19,549,669,0244.639—bartowski
I1-Q3_K_L18.21 GiB19,549,669,5044.639—mradermacher
Q4_K_M19.92 GiB21,383,628,1605.074—ibm-granite
Q4_K_M19.92 GiB21,383,628,4485.074—bartowski
I1-Q4_K_M19.92 GiB21,383,628,9285.074—mradermacher
Q5_K_S21.80 GiB23,407,904,4165.554—bartowski
I1-Q5_K_S21.80 GiB23,407,904,8965.554—mradermacher
Q5_K_M23.05 GiB24,749,852,3205.873—bartowski

Compare with

same modality, comparable size
vs Qwen3-Coder-30B-A3B-Instructvs Qwen3.6-27Bvs Qwen3.8-27Bvs gemma-4-12B-it-qat-q4_0-unquantizedvs gemma-4-E4B-itvs Qwen3-30B-A3B-Thinking-2507All 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 17.66 GiB. The real file is 19.92 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from config.json
Layers
88
Attention heads
—
KV heads
—
Head dim
—
Hidden size
—
Vocab
49,152
Sliding window
none
SWA period
—
MLA
no
Experts
—
Experts per token
—
use_sliding_window
—

Questions people ask

How much VRAM does granite-34b-code-base-8k need?
Q4_K_M is exactly 21,383,628,160 bytes (19.92 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-34b-code-base-8k 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.