Skip to content
Models / Phi-4 Mini

Phi-4 Mini

edge
HuggingFace → HuggingBay → Microsoft → Cost & VRAM calculator → Find for my hardware

AI-generated content marks

The provider reports that this model does not add embedded watermarks or provenance metadata to generated output.

Provider transparency docs →

coding reasoning
Parameters
3.8B
Context
128k
License
mit
Developer
Microsoft
Origin
🇺🇸 USA
Released
Feb 2025

Scores

Coding
55
Reasoning
58
General
52

Related models

Phi-4 14B
14.7B consumer 9GB min RAM $0.07/M in Dec 12, 2024

Save your hardware and every model page answers the real question: will it run on your machine, and how fast?

Join free - save your rig →

Run it locally

Per-quant memory needs and a static "can you run it?" reference - no rig entry required

Q4_K_M
2.5GB 2.5GB min 3.7GB rec
Balanced - the usual local sweet spot
Q8_0
4.1GB 4.1GB min 5.3GB rec
Near-lossless

The reference hardware

Schematic of the 4x H100 80GB (320GB) reference rig - 320GB VRAM, 13400 GB/s aggregate bandwidth
Schematic of the NVIDIA DGX Station 748GB reference rig - 748GB unified memory, 8000 GB/s aggregate bandwidth
Schematic of the 8x RTX 3090 rack (192GB) reference rig - 192GB VRAM, 7489 GB/s aggregate bandwidth
Schematic of the 4x RTX 5090 (128GB) reference rig - 128GB VRAM, 7168 GB/s aggregate bandwidth
Schematic of the AMD Instinct MI300X (192GB) reference rig - 192GB VRAM, 5324 GB/s aggregate bandwidth
Schematic of the 4x RTX 4090 (96GB) reference rig - 96GB VRAM, 4032 GB/s aggregate bandwidth
Schematic of the 2x RTX 5090 (64GB) reference rig - 64GB VRAM, 3584 GB/s aggregate bandwidth
Schematic of the 2x RTX 3090 (48GB) reference rig - 48GB VRAM, 1872 GB/s aggregate bandwidth
Schematic of the Single RTX 5090 (32GB) reference rig - 32GB VRAM, 1792 GB/s aggregate bandwidth
Schematic of the RTX PRO 6000 Blackwell (96GB) reference rig - 96GB VRAM, 1792 GB/s aggregate bandwidth
Schematic of the Mac Studio M4 Ultra 192GB reference rig - 192GB unified memory, 1092 GB/s aggregate bandwidth
Schematic of the Mac Studio M4 Ultra 512GB reference rig - 512GB unified memory, 1092 GB/s aggregate bandwidth
Schematic of the Single RTX 4090 (24GB) reference rig - 24GB VRAM, 1008 GB/s aggregate bandwidth
Schematic of the MacBook Pro M5 Max 128GB reference rig - 128GB unified memory, 614 GB/s aggregate bandwidth
Schematic of the Single GTX 1080 Ti (11GB) reference rig - 11GB VRAM, 484 GB/s aggregate bandwidth
Schematic of the Dual EPYC 9004 + 768GB DDR5-4800 reference rig - 768GB unified memory, 460 GB/s aggregate bandwidth
Schematic of the DGX Spark 128GB unified reference rig - 128GB unified memory, 273 GB/s aggregate bandwidth
Schematic of the Ryzen AI Max+ 395 128GB reference rig - 128GB unified memory, 256 GB/s aggregate bandwidth
Schematic of the Jetson AGX Orin 64GB reference rig - 64GB unified memory, 204 GB/s aggregate bandwidth
Schematic of the Epyc + 512GB DDR4-3200 + 2x RTX 3090 reference rig - 560GB unified memory, 204 GB/s aggregate bandwidth
Schematic of the Epyc + 512GB DDR4-2400 + 2x RTX 3090 reference rig - 560GB unified memory, 153 GB/s aggregate bandwidth
Schematic of the NVIDIA Jetson Orin NX 16GB reference rig - 16GB unified memory, 102 GB/s aggregate bandwidth

22 reference configs, drawn in-house. Scroll for more.

Can you run it? - reference rigs

Rig Q4_K_M Q8_0
4x H100 80GB (320GB) fast 2219.9t/s fast 1498.0t/s
NVIDIA DGX Station 748GB fast 1325.3t/s fast 894.3t/s
8x RTX 3090 rack (192GB) fast 1240.8t/s fast 837.3t/s
4x RTX 5090 (128GB) fast 1187.5t/s fast 801.3t/s
AMD Instinct MI300X (192GB) fast 882.1t/s fast 595.3t/s
4x RTX 4090 (96GB) fast 668.0t/s fast 450.7t/s
2x RTX 5090 (64GB) fast 593.7t/s fast 400.7t/s
2x RTX 3090 (48GB) fast 310.2t/s fast 209.3t/s
Single RTX 5090 (32GB) fast 296.9t/s fast 200.3t/s
RTX PRO 6000 Blackwell (96GB) fast 296.9t/s fast 200.3t/s
Mac Studio M4 Ultra 192GB fast 197.4t/s fast 133.2t/s
Mac Studio M4 Ultra 512GB fast 197.4t/s fast 133.2t/s
Single RTX 4090 (24GB) fast 167.0t/s fast 112.7t/s
MacBook Pro M5 Max 128GB fast 111.0t/s fast 74.9t/s
Single GTX 1080 Ti (11GB) fast 80.2t/s fast 54.1t/s
Dual EPYC 9004 + 768GB DDR5-4800 fast 76.3t/s fast 51.5t/s
DGX Spark 128GB unified fast 45.2t/s fast 30.5t/s
Ryzen AI Max+ 395 128GB fast 42.4t/s fast 28.6t/s
Jetson AGX Orin 64GB fast 33.9t/s fast 22.9t/s
Epyc + 512GB DDR4-3200 + 2x RTX 3090 fast 33.9t/s fast 22.9t/s
Epyc + 512GB DDR4-2400 + 2x RTX 3090 fast 25.5t/s ok 17.2t/s
NVIDIA Jetson Orin NX 16GB ok 17.0t/s ok 11.5t/s

Fit tiers use the same will-it-run logic as the rig finder. For comfortable fits, the badge reflects decode speed: fast >=20 t/s, ok 8-20 t/s, slow <8 t/s. t/s is a bandwidth estimate, not a measured benchmark.

Download options

Q4_K_M official -5% vs fp16
2.5GB dl 2.5GB min 3.7GB rec
REC RAM vs largest quant
Q8_0 official -1% vs fp16
4.1GB dl 4.1GB min 5.3GB rec
REC RAM vs largest quant

Or run it in the cloud

No per-token API provider pricing tracked for Phi-4 Mini yet. For flagship list prices, see the calculator.

See who runs Microsoft in production →

PRICE HISTORY

Inference cost over time

Data accumulates from the first daily sync - longer ranges populate over time. Prices come from OpenRouter snapshots, not a historical API.

Loading price history...

Tokenstead community

Need a second opinion on this model?

Ask the people running it on real hardware, then post your tok/s and fit notes so the next reader has a data point.

Join the community Share your rig

Something wrong or stale? Send a correction.