[whatmodel.sh](../../README.md) / Phi-4-reasoning 14B # Phi-4-reasoning 14B Microsoft · released 2025-04 · 14.7B params · mit **score 57.7** ±0.9 · #231 of 732 · from 24 benchmark results across 5 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-reasoning 14B among every model on that task or test, worst to best. 
task            vs the others          rank  counts  tests
coding          ▁▁▁▃▅▇█●▅▄▃▁   #148 of 469*     12%      3
chat & writing  ▁▁▁▃▆▄▆█●▇▇▃    #181 of 402     10%      3
math            ▁▁▃▅█▆●█▄▂▁▁    #175 of 431     19%      3
coding · #148 of 469 · 12% of its score Beats [Qwen3-Coder-Next (80B-A3B)](qwen3-coder-next.md), [Gemini 2.5 Pro](gemini-2-5-pro.md), [Gemini 3.1 Flash-Lite](gemini-3-1-flash-lite.md) here, though they rank higher overall. 
test                                result  vs the others in its run    difficulty      counts  why
HumanEval+ (maker)                   92.9%  ▁·▁▁▂▃▄▄█▅▄● beats 141/145  easy · avg 77%     12%  sharp · several runs
LiveCodeBench (Aug 1, 2024–Feb 1,…   53.8%  █··········● beats 1/1                           —  not counted yet: no other model in the …
LiveCodeBench 2024-08 to 2025-01 …   53.8%  █··········● beats 1/1                           —  not counted yet: no other model in the …
chat & writing · #181 of 402 · 10% of its score Beats [Seed-OSS 36B Instruct](seed-oss-36b.md), [Gemma 2 9B IT SimPO](gemma-2-9b-simpo.md), [Nemotron-Cascade-14B-Thinking](nemotron-cascade-14b-thinking.md) here, though they rank higher overall. Trails [LFM2.5 1.2B Instruct](lfm2-5-1-2b-instruct.md), [LFM2.5 1.2B Thinking](lfm2-5-1-2b-thinking.md), [LFM2.5 8B-A1B](lfm2-5-8b-a1b.md) here, though they rank lower overall. 
test                                result  vs the others in its run   difficulty        counts  why
Arena-Hard (version/judge unspeci…   73.3%  ▃▃▂▅▃▆▅█●▅▂▃ beats 22/30   medium · avg 59%    7.5%  at its level · several runs
IFEval strict (prompt/instruction…   83.4%  █·········●█ beats 1/2     easy · avg 67%      1.6%  small run
IFEval (aggregation unspecified) …   83.4%  ▁▂▄▂▂▄●██▅▂▁ beats 64/145  easy · avg 74%      1.3%  typical
math · #175 of 431 · 19% of its score Beats [MiMo-V2.5 (310B-A15B)](mimo-v2-5.md), [MiniMax-M2.1](minimax-m2-1.md), [Mistral Medium 3.5 (128B)](mistral-medium-3-5.md) here, though they rank higher overall. Trails [OpenThinker3 7B](openthinker3-7b.md), [Olmo 3 7B Think](olmo-3-7b-think.md), [OpenThinker2 32B](openthinker2-32b.md) here, though they rank lower overall. 
test                        result  vs the others in its run  difficulty        counts  why
AIME 2025 (maker)            63.1%  █··●··█··▄·▄ beats 2/6    medium · avg 45%    8.4%  sharp · at its level · small run · seve…
AIME 2024 (maker)            74.6%  ▄···▄●·▄▄··▄ beats 3/6    medium · avg 56%    7.4%  sharp · small run · several runs
HMMT February 2025 (maker)   43.8%  ▁▁▁●▁▅▅▃██▁▄ beats 4/31   hard · avg 32%      3.0%  at its level · small run
other tests (9) 
test                                result  vs the others in its run    difficulty        counts  why
GPQA Diamond (maker)                 67.1%  ●··▄··▄···█▄ beats 1/6      hard · avg 43%       20%  sharp · at its level · several runs
Medmarks-V mean win rate (19 veri…  0.4815  ▁▄▃▃▇●█▃▅▁▁▁ beats 15/30    avg 0.418            15%  sharp
MMLU-Pro (maker)                     74.3%  ▁▁·▁▁▁▃▂▄●█▁ beats 100/188  medium · avg 59%    8.6%  at its level · several runs
Medmarks-V weighted mean win rate…   48.1%  ▂▃▅▂▆▃●▅█▅▃▆ beats 14/34    medium · avg 41%    6.7%  blunt · at its level
FlenQA (3K-token subset) (maker)     97.7%  ▄···▄····▄·● beats 3/4      easy · avg 86%      3.9%  sharp · noisy · several runs
Open PT-LLM Leaderboard average (…   67.9%  ▃▁▁▁▁▁▁▁▂●▅▁ beats 142/276  medium · avg 72%    1.6%  at its level · noisy
EuroEval HellaSwag (English), tes…   48.7%  ▃·▁▁▂▁●▅█▇▄▁ beats 25/100   medium · avg 67%    1.2%  sharp · noisy
Omni-MATH (maker)                    76.6%  █·█·█·█●··██ beats 4/6      easy · avg 68%      1.1%  blunt · at its level · small run · seve…
EuroEval Life in the UK (English …   83.5%  ▁▁▁▁▁▁▁▂▄●▄▃ beats 88/146   easy · avg 82%      0.6%  noisy
## How the score was worked out Each result implies a score on its own, after adjusting for how hard that test is and who ran it. The model's score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its *counts* column above; here it is added up by who ran the tests. [Method](../method.md) 
source                                    results  share of the score
Microsoft (submitted / maker-reported)         10   43% ▇▇▇▇▇
microsoft/Phi-4-reasoning model card (s…        9   32% ▇▇▇▇
Medmarks: A Comprehensive Open-Source L…        2   22% ▇▇▇
EuroEval leaderboards                           2    2% ▇
Open Portuguese LLM Leaderboard                 1    2% ▇
prior: pull toward the middle                        0% ▇
weighted average                                   = 57.7
Licences: Microsoft ([CC BY 4.0](../../licenses/report-2504.21318v1.txt)) · microsoft/Phi-4-reasoning model card ([mit](../../licenses/microsoft--Phi-4-reasoning.txt)) · Medmarks: A Comprehensive Open-Source LLM Benchmark Suite for Medical Tasks ([CC BY 4.0](../../licenses/paper-2605.01417v1-CC-BY-4.0.txt)) · EuroEval leaderboards ([MIT](https://github.com/EuroEval/leaderboards/blob/1a77f1d2f8ba1ba05ab73287fb090d76bf41b2e0/LICENSE)) · Open Portuguese LLM Leaderboard ([Apache-2.0](../../licenses/hf-leaderboard-spaces-Apache-2.0.txt))
Not counted (2) 
source                  test                            score  why
microsoft/Phi-4-reaso…  LiveCodeBench (Aug 1, 2024–Fe…  53.8%  too few models have it yet
Microsoft               LiveCodeBench 2024-08 to 2025…  53.8%  too few models have it yet
## How to run it ### Through an API api (Microsoft Azure) · $0.22/M blended · $0.125 in / $0.5 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-reasoning' (GA; not Phi-4-reasoning-plus), 32K context: input $0.125, output $0.50, 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-reasoning/blob/1de18ec97600877ce63dbf60c73b998da99f0195/config.json): Configured context for Phi-4-reasoning, not Phi-4-reasoning-plus or base Phi-4. Card advertises 32K; configuration resolves exact tokens. ### 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-reasoning 14B (Q4_K_M, needs ~9.1 GB, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/Phi-4-reasoning-GGUF/tree/9afe391a28e1e6496e05d7f8712a1e4eed31d5f2)) ```sh llama cli -hf lmstudio-community/Phi-4-reasoning-GGUF:Q4_K_M ``` 3. Or start a local server and open the address it prints (http://localhost:9931) ```sh llama serve -hf lmstudio-community/Phi-4-reasoning-GGUF:Q4_K_M ``` Other versions: Q8_0 (~15.6 GB) `llama cli -hf lmstudio-community/Phi-4-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-reasoning 14B (Q4_K_M, needs ~9.1 GB, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/Phi-4-reasoning-GGUF/tree/9afe391a28e1e6496e05d7f8712a1e4eed31d5f2)) ```sh lms get https://huggingface.co/lmstudio-community/Phi-4-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 (~15.6 GB) `lms get https://huggingface.co/lmstudio-community/Phi-4-reasoning-GGUF@Q8_0 -y` · 4bit (~8.2 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/Phi-4-reasoning-MLX-4bit -y` · 8bit (~15.6 GB, Mac only) `lms get https://huggingface.co/mlx-community/Phi-4-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-reasoning 14B (Q4_K_M, needs ~9.32 GB, from [ollama.com](https://ollama.com/library/phi4-reasoning/tags)) ```sh ollama run phi4-reasoning:14b ``` Other versions: Q8_0 (~18.14 GB) `ollama run phi4-reasoning:14b-q8_0` ### 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-reasoning 14B (4bit, needs ~8.2 GB, Mac only, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/Phi-4-reasoning-MLX-4bit/tree/83ac684bce36aec035ff73dc155611161104ca67)) ```sh mlx_lm.chat --model lmstudio-community/Phi-4-reasoning-MLX-4bit ``` 3. Or start an OpenAI-compatible server at http://localhost:8080/v1 ```sh mlx_lm.server --model lmstudio-community/Phi-4-reasoning-MLX-4bit ``` Other versions: 8bit (~15.6 GB, Mac only) `mlx_lm.chat --model mlx-community/Phi-4-reasoning-8bit` Model repository and documentation: [huggingface.co/microsoft](https://huggingface.co/microsoft/Phi-4-reasoning) ## Quantization evidence How much of the full-precision result a quantized version kept, measured within one study. Rows from the same study share a table and aren't independent. They don't describe every download of that precision. 
precision                       test                                full → quantized    kept
FP8 W8A8 dynamic (weights sym…  AIME 2025 (lighteval, vLLM, reaso…     0.613 → 0.646  105.4%
FP8 W8A8 dynamic (weights sym…  GPQA Diamond (lighteval, vLLM)         0.646 → 0.665  102.9%
FP8 W8A8 dynamic (weights sym…  MMLU-Pro 0-shot chat (lm-evaluati…     0.765 → 0.768  100.5%
FP8 W8A8 dynamic (weights sym…  MATH-500 (lighteval, vLLM)               0.9 → 0.886   98.4%
Each kept value links to its source. [interactive version](https://whatmodel.sh/model/phi-4-reasoning) · [JSON](https://whatmodel.sh/api?format=json&all=1)