RunsOnMy ModelsGPUs & MacsBy VRAMBest GPUCan I run…?OllamaStart hereAPI pricesVRAM calculatorNew

Can I run Phi 4 Mini Instruct locally?

Phi 4 Mini Instruct by microsoft has 3.8B parameters. At 4-bit with an 8K context it needs about 4.1 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.

4.1 GBQ4_K_M · 8K context
RTX 4060smallest consumer GPU that fits
M1 16GBsmallest 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 bartowski/microsoft_Phi-4-mini-instruct-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx32K ctx128K ctx
2-bit (Q2_K) •rough2.8 GB3.3 GB6.3 GB18.3 GB
3-bit (Q3_K_M) •usable3.3 GB3.8 GB6.8 GB18.8 GB
4-bit (Q4_K_M) •good, most popular3.6 GB4.1 GB7.1 GB19.1 GB
5-bit (Q5_K_M) •very good3.9 GB4.4 GB7.4 GB19.4 GB
6-bit (Q6_K) •near lossless4.2 GB4.7 GB7.7 GB19.7 GB
8-bit (Q8_0) •effectively lossless5.1 GB5.6 GB8.6 GB20.6 GB
16-bit (F16/BF16)original weights8.4 GB8.9 GB11.9 GB23.9 GB

Phi 4 Mini Instruct on Ollama: download size and VRAM

ollama run phi4-mini:3.8b
SourceFileDownload sizeGPU memory at 8K ctx (est.)
Ollama download sizephi4-mini:3.8b (Q4_K_M)2.5 GB4.1 GB
Our GGUF (Hugging Face)bartowski/microsoft_Phi-4-mini-instruct-GGUF (Q4_K_M)2.5 GB4.1 GB

Download sizes are decimal GB, as ollama.com lists them; memory figures on this site use binary GB (GiB). GPU memory for the Ollama tag is its file size plus this model's KV cache at 8K context and runtime overhead; Ollama's default context window may be smaller (set num_ctx to change it). Source: ollama.com/library/phi4-mini:3.8b, checked Nov 17, 2026. Refreshed weekly; shown only when the tag's architecture and parameter count match this model. All Ollama models with sizes and VRAM.

Which GPUs and Macs run Phi 4 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.)Details
RTX 3060 12GB12 GBFitsF168.9 GB47details
RTX 4060 Ti 16GB16 GBFitsF168.9 GB38details
RTX 309024 GBFitsF168.9 GB122details
RTX 409024 GBFitsF168.9 GB131details
RTX 509032 GBFitsF168.9 GB234details
RTX 5070 Ti16 GBFitsF168.9 GB117details
RX 7900 XTX24 GBFitsF168.9 GB125details
Mac M4 Pro 48GB48 GBFitsF168.9 GB36details
Mac M4 Max 128GB128 GBFitsF168.9 GB71details
Mac M3 Ultra 512GB512 GBFitsF168.9 GB107details
DGX Spark128 GBFitsF168.9 GB36details
RTX 407012 GBFitsF168.9 GB66details
RTX 5060 Ti 16GB16 GBFitsF168.9 GB58details
RTX 40608 GBFitsQ8_05.6 GB67details
Arc B58012 GBFitsF168.9 GB59details
RTX 3060 Ti8 GBFitsQ8_05.6 GB110details
RTX 30708 GBFitsQ8_05.6 GB110details
RTX 3080 10GB10 GBFitsQ8_05.6 GB186details
RTX 3080 12GB12 GBFitsF168.9 GB119details
RTX 3090 Ti24 GBFitsF168.9 GB131details
RTX 4060 Ti 8GB8 GBFitsQ8_05.6 GB71details
RTX 4070 Super12 GBFitsF168.9 GB66details
RTX 4070 Ti12 GBFitsF168.9 GB66details
RTX 4070 Ti Super16 GBFitsF168.9 GB88details
RTX 408016 GBFitsF168.9 GB93details
RTX 4080 Super16 GBFitsF168.9 GB96details
RTX 50608 GBFitsQ8_05.6 GB110details
RTX 507012 GBFitsF168.9 GB88details
RTX 508016 GBFitsF168.9 GB125details
2x RTX 30902×24 GBFitsF168.9 GB122details
2x RTX 40902×24 GBFitsF168.9 GB131details
2x RTX 50902×32 GBFitsF168.9 GB234details
RTX A600048 GBFitsF168.9 GB100details
RTX 6000 Ada48 GBFitsF168.9 GB125details
RTX PRO 6000 Blackwell96 GBFitsF168.9 GB234details
L40S48 GBFitsF168.9 GB113details
A100 40GB40 GBFitsF168.9 GB203details
A100 80GB80 GBFitsF168.9 GB266details
H100 PCIe80 GBFitsF168.9 GB261details
H100 SXM80 GBFitsF168.9 GB437details
H200141 GBFitsF168.9 GB626details
RX 76008 GBFitsQ8_05.6 GB71details
RX 7700 XT12 GBFitsF168.9 GB56details
RX 7800 XT16 GBFitsF168.9 GB81details
RX 7900 GRE16 GBFitsF168.9 GB75details
RX 7900 XT20 GBFitsF168.9 GB104details
RX 907016 GBFitsF168.9 GB83details
RX 9070 XT16 GBFitsF168.9 GB83details
Ryzen AI Max+ 395 128GB128 GBFitsF168.9 GB33details
Arc A770 16GB16 GBFitsF168.9 GB73details
Mac M1 16GB16 GBFitsF168.9 GB8.9details
Mac M2 24GB24 GBFitsF168.9 GB13details
Mac M3 24GB24 GBFitsF168.9 GB13details
Mac M4 16GB16 GBFitsF168.9 GB16details
Mac M4 32GB32 GBFitsF168.9 GB16details
Mac M1 Pro 32GB32 GBFitsF168.9 GB26details
Mac M3 Pro 36GB36 GBFitsF168.9 GB20details
Mac M4 Pro 24GB24 GBFitsF168.9 GB36details
Mac M4 Pro 64GB64 GBFitsF168.9 GB36details
Mac M1 Max 64GB64 GBFitsF168.9 GB52details
Mac M2 Max 96GB96 GBFitsF168.9 GB52details
Mac M3 Max 128GB128 GBFitsF168.9 GB52details
Mac M4 Max 64GB64 GBFitsF168.9 GB71details
Mac M2 Ultra 192GB192 GBFitsF168.9 GB104details
Mac M3 Ultra 256GB256 GBFitsF168.9 GB107details

Questions

How much VRAM does Phi 4 Mini Instruct need?

About 4.1 GB at 4-bit (Q4_K_M) with an 8K context: 2.3 GB of weights (real GGUF file size), 1.0 GB of KV cache and 0.8 GB of runtime overhead. At 8-bit it needs about 5.6 GB.

What is the smallest GPU that runs Phi 4 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).

How big is the Ollama download for Phi 4 Mini Instruct?

2.5GB for the phi4-mini:3.8b tag (Q4_K_M), as listed on ollama.com. Run it with "ollama run phi4-mini:3.8b". With an 8K context, plan on about 4.1 GB of GPU memory.

Can I run Phi 4 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.

Similar models

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

Phi 3 Mini 4K Instruct

3.8B · 5.7 GB at 4-bit

Phi 3.5 Mini Instruct

3.8B · 5.7 GB at 4-bit

Phi 3 Mini 128K Instruct

3.8B · 5.7 GB at 4-bit

Qwen2.5 VL 3B Instruct

3.8B · 2.8 GB at 4-bit

Granite Vision 4.1 4B

4.0B · 3.2 GB at 4-bit

Granite 4.2 3B

3.7B · 3.3 GB at 4-bit

Embed this

Phi 4 Mini Instruct: runs on 8 GB+ VRAM Free badge and "Will it run on my GPU?" widget for model cards, READMEs and blogs. No cookies, no tracking. All options.

Copy a snippet

Markdown badge

[![Phi 4 Mini Instruct: runs on 8 GB+ VRAM (RunsOnMy)](https://runsonmy.com/badge/phi-4-mini-instruct.svg)](https://runsonmy.com/model/phi-4-mini-instruct/)

Widget (script)

<div class="runsonmy-widget" data-model="phi-4-mini-instruct"><a href="https://runsonmy.com/model/phi-4-mini-instruct/">Will Phi 4 Mini Instruct run on my GPU? Check on RunsOnMy</a></div>
<script async src="https://runsonmy.com/embed.js"></script>

Widget (iframe)

<iframe src="https://runsonmy.com/embed/m/phi-4-mini-instruct/" width="100%" height="220" style="max-width:440px;border:0" loading="lazy" title="Will Phi 4 Mini Instruct run on my GPU? (RunsOnMy)"></iframe>

Model facts

Hugging Face repomicrosoft/Phi-4-mini-instruct
Parameters3,836,021,760 (3.8B)
Layers / hidden size32 / 3072
AttentionGQA; 24 heads, 8 KV heads, head dim 128
KV cache per token (FP16)128.0 KiB
Max context131,072 tokens
Licensemit
Released on HFFeb 19, 2025
Downloads (30d) / likes301,567 / 891

Sources: Hugging Face model API · config.json (microsoft/Phi-4-mini-instruct) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.