Phi-4 (14B) on Orange Pi AIpro (20T) (12 GB)

Can the Orange Pi AIpro (20T) (12 GB) run Phi-4 (14B)?

Barely: it runs very slowly (tight fit)

Very slow (tight fit). At Q2_K it needs about 8.9 GB of 9.8 GB usable, with an estimated speed of 1.9 tok/s.

Best quant: Q2_K
Generation speed: 1.9 tok/s
Prompt processing: 2.0 tok/s
First token (1K prompt): 501.8 s
Largest safe context: 12K

Best app to run it on the Orange Pi AIpro (20T) (12 GB)

On Linux on ARM boards

Ollama

Command line + local API · runs on llama.cpp (ggml), MLX (Apple silicon, preview)

One command with Ollama’s own build of this model.
Serves an OpenAI-compatible API that other apps and coding tools can use.

ollama run phi4:14b

LM Studio Point-and-click: search, download and chat in one window.
llama.cpp Usually the first to support new model architectures.

How we choose: the files published for this model, your operating system and GPU, whether it fits fully in memory, and published speed tests (MLX vs llama.cpp on Apple silicon, vLLM vs llama.cpp vs Ollama on NVIDIA). Details.

Download Phi-4 (14B)

Recommended for the Orange Pi AIpro (20T) (12 GB): Q2_K. Every other quantization has its own download link in the table below.

Quickest · Ollama

One command downloads a ready-to-run version and starts a chat.

Get Ollama at ollama.com/download (opens in a new tab) · model page (opens in a new tab)

ollama run phi4:14b

Quantized files · GGUF

Works in LM Studio, llama.cpp, Jan and Ollama. No sign-in needed.

Download the Q2_K file (opens in a new tab)

huggingface.co/microsoft/phi-4-gguf (opens in a new tab)

ollama run hf.co/microsoft/phi-4-gguf:Q2_K

Original weights · Hugging Face

Full-precision files from Microsoft, for fine-tuning, vLLM, Transformers.

huggingface.co/microsoft/phi-4 (opens in a new tab)

No sign-in needed.

Official repo · GitHub

Reference code, inference scripts and docs from Microsoft. The weights themselves download from Hugging Face.

github.com/microsoft/PhiCookBook (opens in a new tab)

Microsoft’s official Phi cookbook: setup, inference and fine-tuning guides.

git clone https://github.com/microsoft/PhiCookBook

Links checked 3 days ago against Hugging Face, Ollama and GitHub.

How it’s rated

Quality 44Fair quality (C) · 44 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 62/100 (quality, fit, speed and memory use) sets the ranking.: How capable the model itself is at Q2_K, from its editorial quality tier scaled by the quantization or precision. It’s the same on any hardware that runs this variant. Scale: A+ 85+ · A 70+ · B 55+ · C 40+ · D under 40.
Runs Very slowtight fit: How well it runs on the Orange Pi AIpro (20T) (12 GB): speed, capped by how well it fits in memory. Runs great: 30+ tokens/s (or near the image, video or decision speed target) · Runs well: 10+ tokens/s · Slow: 3+ tokens/s · Very slow: under 3 · Won’t run: doesn’t fit even with offload. A tight fit or RAM offload caps it at Runs well; heavy offload caps it at Slow.

Match score

To order lists, each result also gets a match score out of 100 from four parts. It favors the most capable model that still runs well.

Model qualityEditorial quality tier of the model, scaled by the quantization or precision used.	20.2 / 45
Memory fitFull marks when the model fits in the share of unified memory the GPU can use; fewer when it goes past that limit.	24.0 / 25
SpeedGeneration speed against a 40 tokens/s target for LLMs, or render time against a target for image and video.	3.3 / 15
Memory useRewards using the memory you have, so a bigger, better variant beats a tiny one that leaves most memory idle.	15.0 / 15
Match score	62.5 / 100

The match score orders results; it is not shown as a grade. Quality and Runs are the two ratings in the tables.

Every quantization on the Orange Pi AIpro (20T) (12 GB)

Quant	Memory needed	Available	Speed	Runs	Quality	Download
`Q2_K`	8.9 GB	9.8 GB	1.9 tok/s	Very slowtight fit	44Fair quality (C) · 44 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 62/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q3_K_M`	10.3 GB	9.8 GB	1.3 tok/s	Very slowover memory limit	53Fair quality (C) · 53 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 51/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (Q3_K_L) (opens in a new tab)
`Q4_K_M`	12.0 GB	9.8 GB	—	Won't run	60Good quality (B) · 60 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 8/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (Q4_K_S) (opens in a new tab)
`Q5_K_M`	13.5 GB	9.8 GB	—	Won't run	62Good quality (B) · 62 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 8/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (Q5_K_S) (opens in a new tab)
`Q6_K`	15.1 GB	9.8 GB	—	Won't run	63Good quality (B) · 63 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 8/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q8_0`	18.6 GB	9.8 GB	—	Won't run	64Good quality (B) · 64 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 8/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`F16`	32.5 GB	9.8 GB	—	Won't run	64Good quality (B) · 64 of 100: how capable the model is at F16, whatever it runs on. Match score 8/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab) Original weights (opens in a new tab)

Memory figures assume 8K of context. Longer conversations or documents grow the KV cache.

Other hardware that runs it

64Good quality (B) · 64 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 83/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 4090 Q8_0 · 42.9 tok/s · Runs great
64Good quality (B) · 64 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 83/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 3090 Q8_0 · 39.8 tok/s · Runs great
63Good quality (B) · 63 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 83/100 (quality, fit, speed and memory use) sets the ranking. AMD Radeon RX 7900 XTX Q6_K · 47.9 tok/s · Runs great
62Good quality (B) · 62 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 82/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 4070 Ti SUPER Q5_K_M · 46.4 tok/s · Runs great
62Good quality (B) · 62 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 82/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 4080 SUPER Q5_K_M · 50.8 tok/s · Runs great
62Good quality (B) · 62 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 82/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 5070 Ti Q5_K_M · 61.8 tok/s · Runs great

Quick answers

How much memory does Phi-4 (14B) need on the Orange Pi AIpro (20T) 12GB?

At Q2_K it needs about 8.9 GB. The Orange Pi AIpro (20T) 12GB has 9.8 GB usable.

How fast is Phi-4 (14B) on the Orange Pi AIpro (20T) 12GB?

About 1.9 tok/s at Q2_K, an estimate from the hardware’s memory bandwidth and the size of the model.

Which quantization of Phi-4 (14B) should I use on the Orange Pi AIpro (20T) 12GB?

Q2_K. Of the versions that fit, it is the one that runs best; when two run equally well, the check prefers the higher-rated one.

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.