[whatmodel.sh](../../README.md) / Phi-4-mini-reasoning 3.8B
# Phi-4-mini-reasoning 3.8B
Microsoft · released 2025-04 · 3.8B params · mit
**score 46.2** ±2.2 · #449 of 732 · from 6 benchmark results across 3 sources
Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only.
## What it's good at
● marks Phi-4-mini-reasoning 3.8B among every model on that task or test, worst to best.
task vs the others rank counts tests
math ▁▁▃▅█▆●█▄▂▁▁ #181 of 431* 25% 2
math · #181 of 431 · 25% of its score
Beats [MiniMax-M2.1](minimax-m2-1.md), [Mistral Medium 3.5 (128B)](mistral-medium-3-5.md), [GPT-4.1](gpt-4-1.md) here, though they rank higher overall.
test result vs the others in its run difficulty counts why
AIME 2024 (maker) 57.5% ▁▁▁▂▁▃▁●███▃ beats 27/85 medium · avg 54% 20% sharp · at its level · noisy
MATH-500 (maker) 94.6% ▁▃▂▃▄▇▅●█▄▂▁ beats 52/85 very easy · avg 86% 5.0% noisy
other tests (4)
test result vs the others in its run difficulty counts why
GPQA Diamond (maker) 52.0% ▁▁▁▁▂▄●▅███▄ beats 94/322 medium · avg 52% 31% sharp
BRIDGE clinical text, zero-shot a… 21.3% ▁▁▁▁▁●▂▃▄▆█▃ beats 14/97 hard · avg 31% 25% blunt · at its level
EuroEval HellaSwag (English), tes… 41.0% ▃·▁▁▂●▂▅█▇▄▁ beats 22/100 medium · avg 67% 10% sharp · at its level · noisy
EuroEval Life in the UK (English … 55.7% ▁▁▁▁▁▁▁●▄█▄▃ beats 26/146 easy · avg 82% 7.0% at its level · noisy
source results share of the score
microsoft/Phi-4-mini-reasoning model ca… 3 56% ▇▇▇▇▇▇▇
BRIDGE Medical Leaderboard 1 25% ▇▇▇
EuroEval leaderboards 2 17% ▇▇
prior: pull toward the middle 2% ▇
weighted average = 46.2
Licences: microsoft/Phi-4-mini-reasoning model card ([MIT](../../licenses/card-microsoft-phi-4-mini-reasoning.txt)) · BRIDGE Medical Leaderboard ([MIT](../../licenses/bridge-medical-leaderboard-MIT.txt)) · EuroEval leaderboards ([MIT](https://github.com/EuroEval/leaderboards/blob/1a77f1d2f8ba1ba05ab73287fb090d76bf41b2e0/LICENSE))
## How to run it
### Through an API
api (Microsoft Azure) · $0.13/M blended · $0.075 in / $0.3 out per 1M tokens · [azure.microsoft.com](https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/microsoft)
[price source](https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/microsoft): USD per 1M tokens, Azure AI Foundry Models pay-as-you-go (Microsoft models table), 'Phi-4-mini-reasoning' (GA), 128K context: input $0.075, output $0.30, region US East. Currency: USD. Microsoft Azure official pricing page (price cells carry the amounts in the page's own data), observed 2026-09-25. · [context source](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/0e3b1e2d02ee478a3743abe3f629e9c0cb722e0a/README.md): Microsoft official model card: Phi-4-mini-reasoning context length 128K tokens (stated rounded). Observed 2026-09-25.
### llama.cpp (recommended)
1. Install llama.cpp. Homebrew and winget installs name the commands llama-cli and llama-server instead. [Other ways to install](https://github.com/ggml-org/llama.cpp/blob/master/docs/install.md)
```sh
# macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
# Windows
irm https://llama.app/install.ps1 | iex
```
2. Run Phi-4-mini-reasoning 3.8B (Q4_K_M, needs ~2.49 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/Phi-4-mini-reasoning-GGUF/tree/6b9cfad9419cad66996ad42ebddbff9b34446479))
```sh
llama cli -hf unsloth/Phi-4-mini-reasoning-GGUF:Q4_K_M
```
3. Or start a local server and open the address it prints (http://localhost:9931)
```sh
llama serve -hf unsloth/Phi-4-mini-reasoning-GGUF:Q4_K_M
```
Other versions: Q8_0 (~4.08 GB) `llama cli -hf unsloth/Phi-4-mini-reasoning-GGUF:Q8_0`
### LM Studio
1. Install LM Studio. Open the app once before using the lms command. [Download it (macOS / Windows / Linux)](https://lmstudio.ai/download)
2. Run Phi-4-mini-reasoning 3.8B (Q4_K_M, needs ~2.49 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/Phi-4-mini-reasoning-GGUF/tree/6b9cfad9419cad66996ad42ebddbff9b34446479))
```sh
lms get https://huggingface.co/unsloth/Phi-4-mini-reasoning-GGUF@Q4_K_M -y
```
3. Then run lms ls to see its name, and lms chat followed by that name
Other versions: Q8_0 (~4.08 GB) `lms get https://huggingface.co/unsloth/Phi-4-mini-reasoning-GGUF@Q8_0 -y` · 4bit (~2.2 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/Phi-4-mini-reasoning-MLX-4bit -y` · 8bit (~4.1 GB, Mac only) `lms get https://huggingface.co/mlx-community/Phi-4-mini-reasoning-8bit -y`
### Ollama
1. Install Ollama. [Download it (macOS / Windows)](https://ollama.com/download)
```sh
# Linux
curl -fsSL https://ollama.com/install.sh | sh
```
2. Run Phi-4-mini-reasoning 3.8B (Q4_K_M, needs ~3.15 GB, from [ollama.com](https://ollama.com/library/phi4-mini-reasoning/tags))
```sh
ollama run phi4-mini-reasoning:3.8b
```
### MLX (Mac)
1. Install MLX (Mac). Apple silicon Macs only, text only.
```sh
# with uv
uv tool install mlx-lm
# or
pip install mlx-lm
```
2. Run Phi-4-mini-reasoning 3.8B (4bit, needs ~2.2 GB, Mac only, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/Phi-4-mini-reasoning-MLX-4bit/tree/10906e3b888052ad3c7b83ffad45ed576cacc905))
```sh
mlx_lm.chat --model lmstudio-community/Phi-4-mini-reasoning-MLX-4bit
```
3. Or start an OpenAI-compatible server at http://localhost:8080/v1
```sh
mlx_lm.server --model lmstudio-community/Phi-4-mini-reasoning-MLX-4bit
```
Other versions: 8bit (~4.1 GB, Mac only) `mlx_lm.chat --model mlx-community/Phi-4-mini-reasoning-8bit`
Model repository and documentation: [huggingface.co/microsoft](https://huggingface.co/microsoft/Phi-4-mini-reasoning)
[interactive version](https://whatmodel.sh/model/phi-4-mini-reasoning) · [JSON](https://whatmodel.sh/api?format=json&all=1)