Can I run Phi-4 locally?

Yes — the weights are open under an MIT licence and are available on Hugging Face at microsoft/phi-4, so you can run it locally with tools like Ollama or LM Studio. At 14B parameters you will generally want 12 to 16 GB of VRAM for comfortable inference.

What does Phi-4 cost via API?

Microsoft prices it at $0.07 per million input tokens and $0.14 per million output tokens, which puts it at the cheaper end of hosted models; the MIT licence also means you can self-host to avoid per-token costs entirely.

How does Phi-4 differ from larger models?

Phi-4 trades raw scale for data quality: Microsoft trained it on synthetic and curated academic data specifically to punch above its weight on reasoning tasks, making it a practical choice when you cannot afford the memory or latency of a 70B-class model.

Phi 4

4099°

#llm #microsoft #reasoning #math #coding #local

Run this model

Offers / Ads

OpenRouter Terminal / API

Pay per token

Chat with this model in the browser or call it from your code with one API key.

Microsoft | 16,384 token context | $0.07/M input, $0.14/M output. [Microsoft Research](/microsoft) Phi-4 is designed to perform well in complex reasoning tasks and can operate efficiently in situations with limited memory or where quick responses are needed. At 14 billion...

Guides for this model

Ollama vs LM Studio (2026): which local AI app to use, and which top models each one runs

Best local LLM by use case (October 2026): the right open model for every job

How much VRAM do I actually need for local AI?

What is a GGUF file? Quantized models explained (Q4, Q8, fp8, NF4)

All guides

Captain's notes

Written by Captain, the house editor

Notes from 2026-10-13

Phi-4 is a 14-billion-parameter text model from Microsoft Research, trained on a carefully curated mix of synthetic data, academic books, and filtered web content, with a deliberate focus on reasoning quality over raw scale. It runs under an MIT licence, so you can download the weights and use them commercially without asking permission. If you need a capable reasoning model that fits on a single consumer GPU, this is one of the more serious options at that size.

Known for

Strong complex reasoning relative to its 14B parameter count
MIT licence: download, modify and deploy commercially, no strings
Trained on 9.8T tokens with heavy emphasis on data quality
Efficient enough for memory-constrained or latency-sensitive deployments
Supervised fine-tuning and direct preference optimisation for instruction following

Watch out for

Context window is 16K tokens — short by current standards; not suited for long documents.
Primarily English; Microsoft flags it is not specifically evaluated for other languages.
Knowledge cutoff is June 2024 or earlier; recent events will not be in its training data.

Best for

Prompts to try

Multi-step logic puzzle

Five people — Alice, Ben, Cara, Dan and Eve — each own exactly one pet: a cat, dog, fish, rabbit or parrot. Alice does not own the fish. Ben owns the parrot. Cara owns the rabbit. Dan does not own the cat. Who owns the fish, and who owns the cat? Show your reasoning step by step.

Refactor for clarity

Refactor this Python function so it is easier to read and handles the case where the input list is empty. Explain each change you make and why it improves the code. def calc(lst): t=0 for i in lst: t=t+i return t/len(lst)

Fermi estimate walkthrough

Estimate how many piano tuners work in a city of two million people. Walk through every assumption you make, assign a rough number to each, and show how they combine into a final estimate. Then say which assumption is most likely to be wrong and in which direction.

Compare with

Space Bunny Alpha
API-only model
1,000,000 token context · free to use
Solar Pro 4
API-only model
524,288 token context · $0.09/M input, $0.36/M output
Gemini 2.5 Flash
API-only model
1,048,576 token context · $0.30/M input, $2.50/M output
Qwen3 30B A3B Thinking 2507
API-only model
81,920 token context · $0.20/M input, $2.40/M output

More text models from microsoft

Questions people ask

Can I run Phi-4 locally?: Yes — the weights are open under an MIT licence and are available on Hugging Face at microsoft/phi-4, so you can run it locally with tools like Ollama or LM Studio. At 14B parameters you will generally want 12 to 16 GB of VRAM for comfortable inference.
What does Phi-4 cost via API?: Microsoft prices it at $0.07 per million input tokens and $0.14 per million output tokens, which puts it at the cheaper end of hosted models; the MIT licence also means you can self-host to avoid per-token costs entirely.
How does Phi-4 differ from larger models?: Phi-4 trades raw scale for data quality: Microsoft trained it on synthetic and curated academic data specifically to punch above its weight on reasoning tasks, making it a practical choice when you cannot afford the memory or latency of a 70B-class model.

Captain's verdict
A genuinely useful small model if reasoning quality matters more to you than context length or multilingual breadth.

Model card on Hugging Face

Captain writes these notes from the model's own description and model card. Prices and context sizes are the provider's at the time of writing and can change.

Ranking badge

Updated Oct 22, 2026

Made this model? Show this badge on your page.

Paste it into your Hugging Face card or Civitai description.

Change colors / personalize / get code

Not this creator?

Get your own badge

How this score adds up

This model has no popularity data from any source yet, so it carries a stand-in number below 7,000° that keeps the lists complete. It never ranks above a model with data. The breakdown appears here as soon as a source reports it.

Glossary: the terms used here

Heat score: One number for how much a model is used and wanted right now, from 0° to 9,999°. Every list on this site that says hottest is sorted by it and nothing else.
Floor: The 7,000° every model with real data starts from. The scale runs from 0° to 9,999°; the part below the floor is reserved for stand-ins (models without data), so that a model with data always ranks above one without. The floor is the same for everyone and decides nothing; the degrees above it are the measurement.
Download-equivalents: Everything the score treats as a download, added up: downloads themselves, each torrent seeder as 1,000, each AI Horde image and each run on a generation site as one.
Log scale: Counting by tens, not by ones: going from 1,000 to 10,000 downloads adds the same degrees as going from 10,000 to 100,000. It keeps giants from crowding out everything else.
Likes: Thumbs up and likes from every source, worth half a download each on the log scale.
Adoption rate: How many downloads a day a model gets right now, from the difference between two readings of a source. A quarter of the weight of the download term.
Measured and estimated: Measured means two readings of a source exist and the difference is real. Estimated means only today's total is known and the past is spread over the model's life; the site says which one it is showing.
Seeder: Someone who keeps a copy of the weights online for others to download over a torrent, or a hosted mirror doing the same. Each one counts as 1,000 downloads.
AI Horde: A volunteer network that generates images for free on donated computers. The number of images it has made with a model is public, and each counts as one download.
Generation sites: Sites that host their own copies of a model and let people generate with it, such as PixAI, SeaArt and Yodayo. Their downloads count like Civitai's and each generation run as one download.
OpenRouter tokens: A text model served through an API, such as Claude or GPT, has no weights to download, so its usage is measured in tokens: the pieces of text it read and wrote through OpenRouter, a gateway many apps use, over the last 30 days. Every 10,000,000 tokens count as one download. Source: OpenRouter (openrouter.ai/rankings), CC BY 4.0.
Full model (checkpoint): The weights themselves, the file that generates on its own. It gets a bonus because add-ons depend on it.
LoRA and embedding: Small add-on files that steer someone else's weights toward a style, a character or a concept. They cannot run alone.
Family and weights: The base a model is built on, such as SD 1.5, SDXL, Flux or Wan. Every model inherits the release date of its family's weights.
Age: Degrees taken off for how long ago the family's weights came out: 150° a year, up to 750°. It keeps old downloads that had years to pile up from freezing the top of the list.
Family boost: A small, bounded nudge for LoRAs of a family this site is pushing forward right now, always listed in the breakdown.
Stand-in score: A stable number below the floor given to a model no source reports on yet, so lists stay complete. It never ranks above a model with data.
New! (grace period): A model less than 7 days old shows New! instead of a score and holds no place in the rankings. Its score is computed from day one, but a model one day old grows faster than anything, so without the wait the week's #1 would always be whatever came out yesterday rather than what held up for a week.
Trend: How a model's growth over a period compares with the typical model's growth over the same period. Faster than typical reads as a rise, slower as a fall.
Typical model: The middle of the catalog: half the models with a figure grew faster over the period, half slower.

Similar Models

Qwen3.8 27B

TEXT / LLM

nomic embed text v1.5

AI Models for Image, Video & Text Generation

Phi 4 is a text model for the Text / LLM family listed on ModelTrends.ai, a read-only catalog of open source AI image, video and text models. Compare it with other Text / LLM models and check its heat score to see how it is trending.

More Text / LLM models llm microsoft reasoning math coding local budget

Browse by Model Family