whatmodel.sh / Phi-4-reasoning-plus 14B

Phi-4-reasoning-plus 14B

+ contribute

Phi-4-reasoning-plus 14B is an open-weights language model from Microsoft, released April 2025. Its published size is 14.7B parameters. Licence: mit.

Catalogue identity notes

Comparison-only, proposed so the Phi-4-reasoning card's 'Model Quality' runs count. The pinned Phi-4-reasoning card (@1de18ec) evaluates two own checkpoints side by side: 'Phi-4-reasoning' (catalogue phi-4-reasoning) and 'Phi-4-reasoning-plus' (additional outcome-based RL on top of Phi-4-reasoning). Weights: https://huggingface.co/microsoft/Phi-4-reasoning-plus (HF API sha 69baf8528e1bcf05f475034d9e5dd32875ed125f, created 2025-04-17, licence mit, 14,659,507,200 parameters in safetensors metadata). Distinct from phi-4-reasoning; do not merge.

58.7±0.8#210 of 719 LLMsfrom 32 benchmark results across 5 sources

$ how it compares

The LLMs ranked either side of Phi-4-reasoning-plus 14B, out of 828 with a score, then more from Microsoft.

#	model	score	where	RAM	released
250	GLM-4.7-Flash Z.ai	59±1	api $0.0000/M +local	~19GB	2026-01
251	EXAONE 4.0 32B LG AI Research	59±1	local	~20GB	2025-07
252	Phi-4-reasoning-plus 14B Microsoft	59±1	local	~10GB	2025-04
253	Nemotron-Cascade-8B NVIDIA	59±1	local	~6.3GB	2025-12
254	Qwen3 14B Alibaba	59±1	api $0.61/M +local	~10GB	2025-04
	more from Microsoft
261	Phi-4-reasoning 14B Microsoft	58±1	api $0.22/M +local	~10GB	2025-04
365	Phi-4 14B Microsoft	53±1	api $0.087/M +local	~10GB	2024-12
422	WizardLM-2 8x22B Microsoft	50±1	local	~91GB	2024-04

see all LLMs · find the ones that fit your machine

$ what it's good at

coding▁▁▁▃▆██▇▆▄▃▁ #172 of 458 · 13% of its score · 5 tests

Beats Qwen3-Coder-Next (80B-A3B), Qwen3.5 9B, Devstral 2 (2512) here, though they rank higher overall. Trails QwQ-32B-Preview, Ministral 3 8B Reasoning 2512, Phi-4-reasoning 14B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
HumanEval+ lab +1	92.3% ✎	▁·▁▁▂▃▄▄█▅▄▃ beats 139/145	easy · avg 78%	▇▇▇ 8.5%	sharp · several runs
LiveCodeBench v6 (window unspecified)	53.1% ✎	█·▄·▄····▄·▄ beats 1/5	medium · avg 42%	▇▇ 3.9%	at its level
SciCode main problems	7.2% ✎	▄··▄····█▄·▄ beats 4/5	very hard · avg 3.8%	▇ 1.0%	too hard for it
LiveCodeBench (Aug 1, 2024–Feb 1, 2025) lab	53.1% ✎	█··········█ beats 0/1		—	not counted yet: no other model in the same run to compare it with
LiveCodeBench 2024-08 to 2025-01 (version not stated) lab	53.1% ✎	█··········█ beats 0/1		—	not counted yet: no other model in the same run to compare it with

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #150 of 393 · 7.5% of its score · 3 tests

Beats INTELLECT-3, MiniMax-M2, MiniCPM5 2B here, though they rank higher overall. Trails LFM2.5 1.2B Thinking, LFM2.5 8B-A1B, LFM2.5 2.6B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Arena-Hard (version/judge unspecified) lab +1	79.0% ✎	▃▃▂▅▃▆▅█▅▅▂▃ beats 27/30	medium · avg 60%	▇▇ 5.3%	several runs
IFEval strict (prompt/instruction level unspecified) lab	84.9% ✎	█·········██ beats 2/2	easy · avg 67%	▇ 1.3%	small run
IFEval (aggregation unspecified) lab	84.9% ✎	▁▁▃▁▂▄▄▆█▃▂▁ beats 73/141	easy · avg 75%	▇ 0.9%	typical

math▁▂▃▅▆▅▄█▄▁▁▁ #158 of 398 · 36% of its score · 5 tests

Beats MiMo-V2.5 (310B-A15B), MiniMax-M2.7, MiniMax-M2.1 here, though they rank higher overall. Trails Olmo 3 7B Think, Qwen3-VL 4B Thinking, Granite 4.2 3B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
AIME 2024 lab +2	81.3% ✎	▄···▄█·▄▄··▄ beats 5/6	medium · avg 62%	▇▇▇▇▇ 12%	sharp · several runs
AIME 2025 lab +2	78.0% ✎	█··▄··█··▄·▄ beats 5/6	medium · avg 53%	▇▇▇▇▇ 12%	sharp · several runs
HMMT 2025 (edition unspecified)	47.7% ✎	▄··▄▄·▄····█ beats 1/5	hard · avg 29%	▇▇▇ 6.9%	sharp · at its level
MATH-500	95.4% ✎	▄·▄·····█··█ beats 1/5	very easy · avg 92%	▇ 2.9%	maxed out
HMMT February 2025 lab	53.6% ✎	▁▁▁▃▃▅▅▃██▁▄ beats 5/31	hard · avg 33%	▇ 2.5%	at its level · small run

other tests9 not in a task · 42% of its score

test	result	vs the others in its run	difficulty	counts	why
GPQA Diamond lab +2	69.3% ✎	█··▄··▄···█▄ beats 2/6	hard · avg 44%	▇▇▇▇▇▇▇▇▇ 22%	sharp · at its level · several runs
MMLU-Pro +2	79.2% ✎	█··█··███··█ beats 5/5	medium · avg 57%	▇▇▇▇ 9.4%	at its level · several runs
Humanity's Last Exam	5.9% ✎	█··██···█·██ beats 2/5	very hard · avg 3.8%	▇▇ 3.8%	typical
FlenQA (3K-token subset) lab +1	97.9% ✎	▄···▄····▄·█ beats 4/4	easy · avg 86%	▇ 3.3%	sharp · noisy · several runs
IFBench (aggregation unspecified)	51.7% ✎	███···██···█ beats 3/5	hard · avg 35%	▇ 1.1%	at its level
Open PT-LLM Leaderboard average (9 Portuguese tasks)	66.0% ✎	▃▁▁▁▁▁▁▁▂█▅▁ beats 121/276	medium · avg 73%	▇ 0.9%	at its level · noisy
Omni-MATH lab +1	81.9% ✎	█·█·█·██··██ beats 5/6	easy · avg 68%	▇ 0.8%	blunt · small run · several runs
EuroEval HellaSwag (English), test split, 5-shot generative MC, accuracy	25.0% ✎	▃·▁▁▂▁▂▅█▇▄▁ beats 0/100	medium · avg 68%	▇ 0.7%	noisy

1 more test

test	result	vs the others in its run	difficulty	counts	why
EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy	52.8% ✎	▁▁▁▁▁▁▃▂▄█▄▃ beats 22/146	easy · avg 83%	▇ 0.4%	noisy

$ where to run it

llama.cpp · recommended

version

Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
Windows
irm https://llama.app/install.ps1 | iex
other ways to install ↗
Run Phi-4-reasoning-plus 14B · needs ~9.1 GB · from huggingface.co/lmstudio-community
chat in the terminal
llama cli -hf lmstudio-community/Phi-4-reasoning-plus-GGUF:Q4_K_M
Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf lmstudio-community/Phi-4-reasoning-plus-GGUF:Q4_K_M

LM Studio

version

Install LM Studio · Open the app once before using the lms command.
download it (macOS / Windows / Linux) ↗
Run Phi-4-reasoning-plus 14B · needs ~9.1 GB · from huggingface.co/lmstudio-community
open in LM Studio ↗
Or download it in the terminal
lms get https://huggingface.co/lmstudio-community/Phi-4-reasoning-plus-GGUF@Q4_K_M -y
Then run lms ls to see its name, and lms chat followed by that name

Ollama

Install Ollama
download it (macOS / Windows) ↗
Linux
curl -fsSL https://ollama.com/install.sh | sh
Run Phi-4-reasoning-plus 14B · needs ~11 GB · from ollama.com
chat in the terminal
ollama run phi4-reasoning:plus

MLX (Mac)

version

Install MLX (Mac) · Apple silicon Macs only, text only.
with uv
uv tool install mlx-lm
or
pip install mlx-lm
Run Phi-4-reasoning-plus 14B · needs ~8.2 GB, Mac only · from huggingface.co/lmstudio-community
chat in the terminal
mlx_lm.chat --model lmstudio-community/Phi-4-reasoning-plus-MLX-4bit
Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model lmstudio-community/Phi-4-reasoning-plus-MLX-4bit

Official weights and documentation: huggingface.co/microsoft ↗

$ how the score was worked out · Falcon-H1R: Pushing the Reaso… 44%, Microsoft — Phi-4-reasoning T… 31%, microsoft/Phi-4-reasoning mod… 23% and 2 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The 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's added up by who ran the tests. How it works

source	results	share of the score	licence
Falcon-H1R: Pushing the Reasoning Frontiers with a Hybrid Model for Efficient Test-Time Scaling	10	▇▇▇▇▇▇▇▇▇▇▇▇▇ 44%	CC BY 4.0
Microsoft — Phi-4-reasoning Technical Report · lab-reported	10	▇▇▇▇▇▇▇▇▇ 31%	CC BY 4.0
microsoft/Phi-4-reasoning model card · lab-reported	9	▇▇▇▇▇▇▇ 23%	mit
EuroEval leaderboards (results collection), English generative leaderboard	2	▇ 1%	MIT
Open Portuguese LLM Leaderboard (archived)	1	▇ 1%	Apache-2.0
prior · pull toward the middle		▇ 0%
weighted average		= 58.7

not counted (2)

microsoft/Phi-4-reasoning model card	LiveCodeBench (Aug 1, 2024–Feb 1, 2025) [correct]	53.1%	fewer than three models have this test yet, or this source ran it for this model alone
Microsoft — Phi-4-reasoning Technical Report	LiveCodeBench 2024-08 to 2025-01 (version not stated) [correct]	53.1%	fewer than three models have this test yet, or this source ran it for this model alone

$ improve this page

Update model information Add a place to run it Add a test result Propose a benchmark Correct a benchmark score Edit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated 2026-10-11 02:20 UTC · sources and licences · json