RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can I run gemma 3 4b pt locally?

gemma 3 4b pt by google has 4.3B parameters. At 4-bit with an 8K context it needs about 3.6 GB of GPU memory. It fits on 65 of 65 GPUs and Macs we track. See all models that fit in 8 GB of VRAM.

3.6 GBQ4_K_M · 8K context
RTX 4060smallest consumer GPU that fits
M1 16GBsmallest Mac that fits
128Kmax context
GQA + sliding windowattention type

GGUF Q8_0 file was 346% of the size expected from the parameter count, so this page uses size estimates instead of real file sizes.

Memory needed by quant and context length

GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx32K ctx128K ctx
2-bit (Q2_K)rough2.6 GB2.7 GB3.1 GB5.0 GB
3-bit (Q3_K_M)usable3.0 GB3.1 GB3.6 GB5.5 GB
4-bit (Q4_K_M)good, most popular3.5 GB3.6 GB4.1 GB6.0 GB
5-bit (Q5_K_M)very good3.9 GB4.0 GB4.5 GB6.4 GB
6-bit (Q6_K)near lossless4.4 GB4.5 GB4.9 GB6.8 GB
8-bit (Q8_0)effectively lossless5.3 GB5.4 GB5.9 GB7.8 GB
16-bit (F16/BF16)original weights9.1 GB9.2 GB9.6 GB11.5 GB

Which GPUs and Macs run gemma 3 4b pt?

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 GBFitsF169.2 GB42details
RTX 4060 Ti 16GB16 GBFitsF169.2 GB33details
RTX 309024 GBFitsF169.2 GB109details
RTX 409024 GBFitsF169.2 GB117details
RTX 509032 GBFitsF169.2 GB208details
RTX 5070 Ti16 GBFitsF169.2 GB104details
RX 7900 XTX24 GBFitsF169.2 GB112details
Mac M4 Pro 48GB48 GBFitsF169.2 GB32details
Mac M4 Max 128GB128 GBFitsF169.2 GB63details
Mac M3 Ultra 512GB512 GBFitsF169.2 GB95details
DGX Spark128 GBFitsF169.2 GB32details
RTX 407012 GBFitsF169.2 GB59details
RTX 5060 Ti 16GB16 GBFitsF169.2 GB52details
RTX 40608 GBFitsQ8_05.4 GB60details
Arc B58012 GBFitsF169.2 GB53details
RTX 3060 Ti8 GBFitsQ8_05.4 GB98details
RTX 30708 GBFitsQ8_05.4 GB98details
RTX 3080 10GB10 GBFitsQ8_05.4 GB166details
RTX 3080 12GB12 GBFitsF169.2 GB106details
RTX 3090 Ti24 GBFitsF169.2 GB117details
RTX 4060 Ti 8GB8 GBFitsQ8_05.4 GB63details
RTX 4070 Super12 GBFitsF169.2 GB59details
RTX 4070 Ti12 GBFitsF169.2 GB59details
RTX 4070 Ti Super16 GBFitsF169.2 GB78details
RTX 408016 GBFitsF169.2 GB83details
RTX 4080 Super16 GBFitsF169.2 GB86details
RTX 50608 GBFitsQ8_05.4 GB98details
RTX 507012 GBFitsF169.2 GB78details
RTX 508016 GBFitsF169.2 GB112details
2x RTX 30902×24 GBFitsF169.2 GB109details
2x RTX 40902×24 GBFitsF169.2 GB117details
2x RTX 50902×32 GBFitsF169.2 GB208details
RTX A600048 GBFitsF169.2 GB89details
RTX 6000 Ada48 GBFitsF169.2 GB112details
RTX PRO 6000 Blackwell96 GBFitsF169.2 GB208details
L40S48 GBFitsF169.2 GB100details
A100 40GB40 GBFitsF169.2 GB181details
A100 80GB80 GBFitsF169.2 GB237details
H100 PCIe80 GBFitsF169.2 GB233details
H100 SXM80 GBFitsF169.2 GB390details
H200141 GBFitsF169.2 GB558details
RX 76008 GBFitsQ8_05.4 GB63details
RX 7700 XT12 GBFitsF169.2 GB50details
RX 7800 XT16 GBFitsF169.2 GB73details
RX 7900 GRE16 GBFitsF169.2 GB67details
RX 7900 XT20 GBFitsF169.2 GB93details
RX 907016 GBFitsF169.2 GB74details
RX 9070 XT16 GBFitsF169.2 GB74details
Ryzen AI Max+ 395 128GB128 GBFitsF169.2 GB30details
Arc A770 16GB16 GBFitsF169.2 GB65details
Mac M1 16GB16 GBFitsF169.2 GB7.9details
Mac M2 24GB24 GBFitsF169.2 GB12details
Mac M3 24GB24 GBFitsF169.2 GB12details
Mac M4 16GB16 GBFitsF169.2 GB14details
Mac M4 32GB32 GBFitsF169.2 GB14details
Mac M1 Pro 32GB32 GBFitsF169.2 GB23details
Mac M3 Pro 36GB36 GBFitsF169.2 GB17details
Mac M4 Pro 24GB24 GBFitsF169.2 GB32details
Mac M4 Pro 64GB64 GBFitsF169.2 GB32details
Mac M1 Max 64GB64 GBFitsF169.2 GB47details
Mac M2 Max 96GB96 GBFitsF169.2 GB47details
Mac M3 Max 128GB128 GBFitsF169.2 GB47details
Mac M4 Max 64GB64 GBFitsF169.2 GB63details
Mac M2 Ultra 192GB192 GBFitsF169.2 GB93details
Mac M3 Ultra 256GB256 GBFitsF169.2 GB95details

Questions

How much VRAM does gemma 3 4b pt need?

About 3.6 GB at 4-bit (Q4_K_M) with an 8K context: 2.4 GB of weights (estimated), 0.3 GB of KV cache and 0.9 GB of runtime overhead. At 8-bit it needs about 5.4 GB.

What is the smallest GPU that runs gemma 3 4b pt?

At Q4_K_M and 8K context, the smallest single consumer GPU in our table that fits it is the RTX 4060 (8 GB).

Can I run gemma 3 4b pt on a Mac?

Yes. At Q4_K_M with 8K context it fits on a Mac M1 16GB 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.

gemma 3 4b it

4.3B · 3.5 GB at 4-bit

medgemma 1.5 4b it

4.3B · 3.5 GB at 4-bit

medgemma 4b it

4.3B · 3.5 GB at 4-bit

Qwen3 VL 4B Instruct

4.4B · 4.1 GB at 4-bit

Phi 3.5 vision instruct

4.1B · 5.8 GB at 4-bit

LightOnOCR 3 4B

4.5B · 3.6 GB at 4-bit

Model facts

Hugging Face repogoogle/gemma-3-4b-pt
Parameters4,300,079,472 (4.3B)
Layers / hidden size34 / 2560
AttentionGQA + sliding window; 8 heads, 4 KV heads, head dim 256; sliding window 1024
KV cache per token (FP16)136.0 KiB (sliding layers capped at the window)
Max context131,072 tokens
Licensegemma (gated: accept terms on Hugging Face)
Released on HFFeb 20, 2025
Downloads (30d) / likes276,894 / 180

Sources: Hugging Face model API · config.json (unsloth/gemma-3-4b-pt) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.