v2.2 · 2026-06-12

Phi-4 Reasoning

phi-4-reasoning

REASONING

Phi-4 Reasoning is a 14B reasoning model profile for local AI planning. This page records provider, license, context, quantization, hardware-fit estimates, setup hints, source links, and caveats so readers do not choose a model by name alone.

PROVIDER

Microsoft

FAMILY

Phi-4

MODEL TYPE

reasoning

PARAMETERS

14B

MODALITIES

text

ARCHITECTURE

Dense decoder-only Transformer; reasoning fine-tune of Phi-4 14B

CONTEXT WINDOW

32,000 tokens

TRAINING TOKENS

—

RELEASE DATE

2025-04-30

LICENSE & LINKS

MITlicense text

MODEL CARD

huggingface.co/microsoft/Phi-4-reasoning

HUGGING FACE

microsoft/Phi-4-reasoning

GGUF REPOSITORIES

bartowski/microsoft_Phi-4-reasoning-GGUF community conversion

Community GGUF repository referenced by the June 2026 model research.

lmstudio-community/Phi-4-reasoning-GGUF community conversion

Community GGUF repository referenced by the June 2026 model research.

SETUP HINTS

OLLAMA

ollama run phi4-reasoning

LM STUDIO

Available according to the supplied June 2026 research; select a cited GGUF/MLX build and verify current app metadata before relying on hands-on setup guidance.

LLAMA.CPP

Use the cited GGUF repository (bartowski/microsoft_Phi-4-reasoning-GGUF) with current llama.cpp-compatible tooling.

QUANTIZATION FILE SIZES (GGUF)

Q4_K_M

9.05 GB

Q5_K_M

10.6 GB

Q8_0

15.58 GB

Quantization note: Q4_K_M for first local attempts unless a cited runtime recommends a different default.

RAM / VRAM ESTIMATES

MIN RAM

16 GB

COMFORTABLE RAM

32 GB

MIN VRAM

16 GB

COMFORTABLE VRAM

32 GB

These are conservative local-inference estimates, not official hardware requirements.

HARDWARE FIT

CPU only (no GPU)

limited

8 GB RAM

16 GB RAM

limited

32 GB RAM

comfortable

8 GB VRAM

12 GB VRAM

24 GB VRAM

limited

Apple Silicon (unified memory)

limited

Hardware fit values are conservative local-inference estimates based on GGUF size plus runtime and KV-cache overhead. Actual requirements depend on context length, quantization, runtime, GPU offload, and other running apps.

BEST FOR

·Harder reasoning, math, science, coding puzzles, and users with enough memory for 14B models locally.

AVOID IF

·Avoid on low-RAM laptops, quick chats, or tasks needing short, direct answers without thinking overhead.

CAVEATS

·This is not the same as base Phi-4 14B. It is heavier than the mini models and may produce long reasoning-style answers. For llama.cpp-style runtimes, verify the correct reasoning chat template and Jinja handling before following runtime-specific instructions.

·Commercial-use status from the supplied research: Permitted under MIT license; verify application-specific obligations before commercial deployment.

·Beginner summary from supplied research: Phi-4 Reasoning is a larger 14B specialty model for difficult math, science, coding, and planning prompts. It is much heavier than mini models, but useful when reasoning quality matters more than speed or memory savings.

SOURCE URLS