RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can I run granite 4.0 h small locally?

granite 4.0 h small by ibm-granite has 32B parameters (8.8B active per token, 72 experts, 10 used per token). At 4-bit with an 8K context it needs about 18.9 GB of GPU memory. It fits on 48 of 65 GPUs and Macs we track. See all models that fit in 24 GB of VRAM.

18.9 GBQ4_K_M · 8K context
RTX 3090smallest consumer GPU that fits
M4 32GBsmallest Mac that fits
128Kmax context
GQAattention type

Memory needed by quant and context length

GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from ibm-granite/granite-4.0-h-small-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx32K ctx128K ctx
2-bit (Q2_K) •rough11.6 GB11.7 GB12.1 GB13.6 GB
3-bit (Q3_K_M) •usable15.0 GB15.0 GB15.4 GB16.9 GB
4-bit (Q4_K_M) •good, most popular18.8 GB18.9 GB19.2 GB20.7 GB
5-bit (Q5_K_M) •very good21.9 GB22.0 GB22.4 GB23.9 GB
6-bit (Q6_K) •near lossless25.3 GB25.4 GB25.7 GB27.2 GB
8-bit (Q8_0) •effectively lossless32.6 GB32.6 GB33.0 GB34.5 GB
16-bit (F16/BF16) •original weights60.7 GB60.7 GB61.1 GB62.6 GB

Which GPUs and Macs run granite 4.0 h small?

Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).

HardwareMemoryFits?Best quantNeedsMax tok/s (est.)
RTX 3060 12GB12 GBNo———details
RTX 4060 Ti 16GB16 GBFitsQ2_K11.7 GB89details
RTX 309024 GBFitsQ4_K_M18.9 GB176details
RTX 409024 GBFitsQ4_K_M18.9 GB189details
RTX 509032 GBFitsQ6_K25.4 GB248details
RTX 5070 Ti16 GBFitsQ2_K11.7 GB278details
RX 7900 XTX24 GBFitsQ4_K_M18.9 GB180details
Mac M4 Pro 48GB48 GBFitsQ6_K25.4 GB38details
Mac M4 Max 128GB128 GBFitsF1660.7 GB31details
Mac M3 Ultra 512GB512 GBFitsF1660.7 GB46details
DGX Spark128 GBFitsF1660.7 GB15details
RTX 407012 GBNo———details
RTX 5060 Ti 16GB16 GBFitsQ2_K11.7 GB139details
RTX 40608 GBNo———details
Arc B58012 GBNo———details
RTX 3060 Ti8 GBNo———details
RTX 30708 GBNo———details
RTX 3080 10GB10 GBNo———details
RTX 3080 12GB12 GBNo———details
RTX 3090 Ti24 GBFitsQ4_K_M18.9 GB189details
RTX 4060 Ti 8GB8 GBNo———details
RTX 4070 Super12 GBNo———details
RTX 4070 Ti12 GBNo———details
RTX 4070 Ti Super16 GBFitsQ2_K11.7 GB209details
RTX 408016 GBFitsQ2_K11.7 GB223details
RTX 4080 Super16 GBFitsQ2_K11.7 GB229details
RTX 50608 GBNo———details
RTX 507012 GBNo———details
RTX 508016 GBFitsQ2_K11.7 GB298details
2x RTX 30902×24 GBFitsQ8_032.6 GB100details
2x RTX 40902×24 GBFitsQ8_032.6 GB108details
2x RTX 50902×32 GBFitsQ8_032.6 GB191details
RTX A600048 GBFitsQ8_032.6 GB82details
RTX 6000 Ada48 GBFitsQ8_032.6 GB103details
RTX PRO 6000 Blackwell96 GBFitsF1660.7 GB102details
L40S48 GBFitsQ8_032.6 GB92details
A100 40GB40 GBFitsQ8_032.6 GB166details
A100 80GB80 GBFitsF1660.7 GB116details
H100 PCIe80 GBFitsF1660.7 GB114details
H100 SXM80 GBFitsF1660.7 GB190details
H200141 GBFitsF1660.7 GB272details
RX 76008 GBNo———details
RX 7700 XT12 GBNo———details
RX 7800 XT16 GBFitsQ2_K11.7 GB194details
RX 7900 GRE16 GBFitsQ2_K11.7 GB179details
RX 7900 XT20 GBFitsQ3_K_M15.0 GB191details
RX 907016 GBFitsQ2_K11.7 GB199details
RX 9070 XT16 GBFitsQ2_K11.7 GB199details
Ryzen AI Max+ 395 128GB128 GBFitsF1660.7 GB15details
Arc A770 16GB16 GBFitsQ2_K11.7 GB174details
Mac M1 16GB16 GBNo———details
Mac M2 24GB24 GBFitsQ2_K11.7 GB31details
Mac M3 24GB24 GBFitsQ2_K11.7 GB31details
Mac M4 16GB16 GBNo———details
Mac M4 32GB32 GBFitsQ4_K_M18.9 GB23details
Mac M1 Pro 32GB32 GBFitsQ4_K_M18.9 GB38details
Mac M3 Pro 36GB36 GBFitsQ4_K_M18.9 GB28details
Mac M4 Pro 24GB24 GBFitsQ2_K11.7 GB85details
Mac M4 Pro 64GB64 GBFitsQ8_032.6 GB29details
Mac M1 Max 64GB64 GBFitsQ8_032.6 GB43details
Mac M2 Max 96GB96 GBFitsF1660.7 GB23details
Mac M3 Max 128GB128 GBFitsF1660.7 GB23details
Mac M4 Max 64GB64 GBFitsQ8_032.6 GB58details
Mac M2 Ultra 192GB192 GBFitsF1660.7 GB45details
Mac M3 Ultra 256GB256 GBFitsF1660.7 GB46details

Questions

How much VRAM does granite 4.0 h small need?

About 18.9 GB at 4-bit (Q4_K_M) with an 8K context: 18.1 GB of weights (real GGUF file size), 0.1 GB of KV cache and 0.6 GB of runtime overhead. At 8-bit it needs about 32.6 GB.

What is the smallest GPU that runs granite 4.0 h small?

At Q4_K_M and 8K context, the smallest single consumer GPU in our table that fits it is the RTX 3090 (24 GB).

Can I run granite 4.0 h small on a Mac?

Yes. At Q4_K_M with 8K context it fits on a Mac M4 32GB or larger, using the default macOS GPU memory limit.

Similar models

Same family or a similar parameter count, with the VRAM each needs at Q4_K_M and 8K context.

granite 4.2 30b

29B · 19.1 GB at 4-bit

granite 4.2 8b

8.8B · 6.8 GB at 4-bit

granite 4.1 8b

8.8B · 6.8 GB at 4-bit

GLM Z1 32B 0414

33B · 19.5 GB at 4-bit

GLM 4 32B 0414

33B · 19.5 GB at 4-bit

Qwen3 32B

33B · 21.1 GB at 4-bit

Model facts

Hugging Face repoibm-granite/granite-4.0-h-small
Parameters32,207,337,984 (32B) · ~8.8B active (computed from config)
Layers / hidden size40 / 4096
AttentionGQA (hybrid: 36 linear/state layers); 32 heads, 8 KV heads, head dim 128
KV cache per token (FP16)16.0 KiB
Max context131,072 tokens
Licenseapache-2.0
Released on HFSep 16, 2025
Downloads (30d) / likes26,257 / 313

Sources: Hugging Face model API · config.json (ibm-granite/granite-4.0-h-small) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.