RunsOnMy ModelsGPUs & MacsAPI pricesVRAM calculatorNew

Can I run Phi 3.5 mini instruct locally?

Phi 3.5 mini instruct by microsoft has 3.8B parameters. At 4-bit with an 8K context it needs about 5.7 GB of GPU memory. It fits on 65 of 65 GPUs and Macs we track.

5.7 GBQ4_K_M · 8K context
RTX 4060smallest consumer GPU that fits
M1 16GBsmallest Mac that fits
128Kmax context
MHAattention type

Memory needed by quant and context length

GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from bartowski/Phi-3.5-mini-instruct-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx32K ctx128K ctx
2-bit (Q2_K) •rough3.3 GB4.8 GB13.8 GB49.8 GB
3-bit (Q3_K_M) •usable3.8 GB5.3 GB14.3 GB50.3 GB
4-bit (Q4_K_M) •good, most popular4.2 GB5.7 GB14.7 GB50.7 GB
5-bit (Q5_K_M) •very good4.6 GB6.1 GB15.1 GB51.1 GB
6-bit (Q6_K) •near lossless4.9 GB6.4 GB15.4 GB51.4 GB
8-bit (Q8_0) •effectively lossless5.7 GB7.2 GB16.2 GB52.2 GB
16-bit (F16/BF16)original weights9.1 GB10.6 GB19.6 GB55.6 GB

Which GPUs and Macs run Phi 3.5 mini instruct?

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 GBFitsQ8_07.2 GB89details
RTX 4060 Ti 16GB16 GBFitsF1610.6 GB38details
RTX 309024 GBFitsF1610.6 GB122details
RTX 409024 GBFitsF1610.6 GB132details
RTX 509032 GBFitsF1610.6 GB234details
RTX 5070 Ti16 GBFitsF1610.6 GB117details
RX 7900 XTX24 GBFitsF1610.6 GB126details
Mac M4 Pro 48GB48 GBFitsF1610.6 GB36details
Mac M4 Max 128GB128 GBFitsF1610.6 GB71details
Mac M3 Ultra 512GB512 GBFitsF1610.6 GB107details
DGX Spark128 GBFitsF1610.6 GB36details
RTX 407012 GBFitsQ8_07.2 GB124details
RTX 5060 Ti 16GB16 GBFitsF1610.6 GB59details
RTX 40608 GBFitsQ6_K6.4 GB87details
Arc B58012 GBFitsQ8_07.2 GB112details
RTX 3060 Ti8 GBFitsQ6_K6.4 GB143details
RTX 30708 GBFitsQ6_K6.4 GB143details
RTX 3080 10GB10 GBFitsQ8_07.2 GB187details
RTX 3080 12GB12 GBFitsQ8_07.2 GB225details
RTX 3090 Ti24 GBFitsF1610.6 GB132details
RTX 4060 Ti 8GB8 GBFitsQ6_K6.4 GB92details
RTX 4070 Super12 GBFitsQ8_07.2 GB124details
RTX 4070 Ti12 GBFitsQ8_07.2 GB124details
RTX 4070 Ti Super16 GBFitsF1610.6 GB88details
RTX 408016 GBFitsF1610.6 GB94details
RTX 4080 Super16 GBFitsF1610.6 GB96details
RTX 50608 GBFitsQ6_K6.4 GB143details
RTX 507012 GBFitsQ8_07.2 GB165details
RTX 508016 GBFitsF1610.6 GB126details
2x RTX 30902×24 GBFitsF1610.6 GB122details
2x RTX 40902×24 GBFitsF1610.6 GB132details
2x RTX 50902×32 GBFitsF1610.6 GB234details
RTX A600048 GBFitsF1610.6 GB100details
RTX 6000 Ada48 GBFitsF1610.6 GB126details
RTX PRO 6000 Blackwell96 GBFitsF1610.6 GB234details
L40S48 GBFitsF1610.6 GB113details
A100 40GB40 GBFitsF1610.6 GB203details
A100 80GB80 GBFitsF1610.6 GB267details
H100 PCIe80 GBFitsF1610.6 GB262details
H100 SXM80 GBFitsF1610.6 GB438details
H200141 GBFitsF1610.6 GB628details
RX 76008 GBFitsQ6_K6.4 GB92details
RX 7700 XT12 GBFitsQ8_07.2 GB106details
RX 7800 XT16 GBFitsF1610.6 GB82details
RX 7900 GRE16 GBFitsF1610.6 GB75details
RX 7900 XT20 GBFitsF1610.6 GB105details
RX 907016 GBFitsF1610.6 GB84details
RX 9070 XT16 GBFitsF1610.6 GB84details
Ryzen AI Max+ 395 128GB128 GBFitsF1610.6 GB33details
Arc A770 16GB16 GBFitsF1610.6 GB73details
Mac M1 16GB16 GBFitsQ8_07.2 GB17details
Mac M2 24GB24 GBFitsF1610.6 GB13details
Mac M3 24GB24 GBFitsF1610.6 GB13details
Mac M4 16GB16 GBFitsQ8_07.2 GB30details
Mac M4 32GB32 GBFitsF1610.6 GB16details
Mac M1 Pro 32GB32 GBFitsF1610.6 GB26details
Mac M3 Pro 36GB36 GBFitsF1610.6 GB20details
Mac M4 Pro 24GB24 GBFitsF1610.6 GB36details
Mac M4 Pro 64GB64 GBFitsF1610.6 GB36details
Mac M1 Max 64GB64 GBFitsF1610.6 GB52details
Mac M2 Max 96GB96 GBFitsF1610.6 GB52details
Mac M3 Max 128GB128 GBFitsF1610.6 GB52details
Mac M4 Max 64GB64 GBFitsF1610.6 GB71details
Mac M2 Ultra 192GB192 GBFitsF1610.6 GB105details
Mac M3 Ultra 256GB256 GBFitsF1610.6 GB107details

Questions

How much VRAM does Phi 3.5 mini instruct need?

About 5.7 GB at 4-bit (Q4_K_M) with an 8K context: 2.2 GB of weights (real GGUF file size), 3.0 GB of KV cache and 0.5 GB of runtime overhead. At 8-bit it needs about 7.2 GB.

What is the smallest GPU that runs Phi 3.5 mini instruct?

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 Phi 3.5 mini instruct 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.

Model facts

Hugging Face repomicrosoft/Phi-3.5-mini-instruct
Parameters3,821,079,552 (3.8B)
Layers / hidden size32 / 3072
AttentionMHA; 32 heads, 32 KV heads, head dim 96
KV cache per token (FP16)384.0 KiB
Max context131,072 tokens
Licensemit
Released on HFAug 16, 2024
Downloads (30d) / likes329,956 / 1,297

Sources: Hugging Face model API · config.json (microsoft/Phi-3.5-mini-instruct) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.