What are the hardware requirements for Phi 4?

Phi 4 requires approximately 29.4 GB of VRAM and supports a context window of 16K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Phi 4 and how large is it?

Phi 4 is developed by microsoft, a model with 14.7b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Phi 4 perform on standard benchmarks?

Phi 4 has the following published scores: LMSYS ELO 1256. Compare against reference models on this page or on the LLM Explorer leaderboards.

Phi 4 by microsoft — VRAM 29.4GB, 16K context

Name: Phi 4
Author: microsoft

Phi 4 is an open-source language model by microsoft. Features: 14.7b LLM, VRAM: 29.4GB, Context: 16K, License: mit, LLM Explorer Score: 0.52, ELO: 1256.

Arxiv:2412.08905 Chat Code Conversational Deploy:azure En Endpoints compatible Eval-results Math Phi Phi3 Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/microsoft/phi-4

Phi 4 Benchmarks

LMSys ELO 1256 vs 1272 (so35) -1.3%

^nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

What is the LLM Explorer Rank (Score)

Phi 4 Parameters and Internals

LLM Name	Phi 4
Repository 🤗	https://huggingface.co/microsoft/phi-4
Model Size	14.7b
Required VRAM	29.4 GB
Updated	2026-08-05
Maintainer	microsoft
Model Type	phi3
Model Files	4.9 GB: 1-of-6 5.0 GB: 2-of-6 4.9 GB: 3-of-6 4.8 GB: 4-of-6 4.8 GB: 5-of-6 5.0 GB: 6-of-6
Supported Languages	en
Model Architecture	Phi3ForCausalLM
License	mit
Context Length	16384
Model Max Length	16384
Transformers Version	4.47.0
Tokenizer Class	GPT2Tokenizer
Padding Token	<\|dummy_85\|>
Vocabulary Size	100352
Torch Data Type	bfloat16

Rank the Phi 4 capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Quantized Models of the Phi 4

Model	Likes	Downloads	VRAM
Phi 4 Unsloth Bnb 4bit	65	8838	10 GB
Phi 4 Bnb 4bit	17	2206	9 GB
Parm 2 CoT 14B 16K O1 QwQ	7	18	29 GB
...perThoughts CoT 14B 16K O1 QwQ	6	265	29 GB
Phi 4 GGUF	8	58229	5 GB
Phi 4 4bit	22	856	8 GB
Phi 4 8bit	13	244	15 GB
Phi 4 6bit	7	50	11 GB

Best Alternatives to Phi 4

Best Alternatives	Context / RAM	Downloads	Likes
Phi 4 Reasoning Plus	32K / 29.4 GB	11601	348
Phi 4 Reasoning	32K / 29.4 GB	20641	228
Phi 4 Reasoning	32K / 29.4 GB	126	2
Phi 4 Reasoning Plus	32K / 29.4 GB	73	7
Flashlight V1.1	16K / 29.4 GB	5	0
Phi 4 RP V0	16K / 29.5 GB	2	3
Phi4 Abliterated	16K / 29.4 GB	15	12
Microsoft Phi 4	16K / 29.4 GB	74	55
Phi 4 Abliterated	16K / 29.4 GB	16	34
DNA R1	16K / 29.4 GB	21	41