Model reference · open weights
phi-4 is an open-weight language model from microsoft. phi-4 (BF16) weighs 29.3 GB; the smallest configuration that runs it is L40S 48 GB.
Summary of the microsoft/phi-4 model card, 2026-10-01
What it is
| Released by | microsoft |
|---|---|
| Released | 2024-12-11 |
| Parameters | 14.7B |
| VRAM | 29.3 GB for the weights |
What it runs on
| Card | Requests at once | Context max | Memory |
|---|---|---|---|
| 8K each | |||
| RTX 3060 12 GB … RTX 4090 24 GB 4 smaller cards | — | — | |
| RTX 5090 32 GB | — | 4K | 31.0 GB |
| L40S 48 GB | 8 | all 16K | 44.0 GB |
| A100 80 GB | 28 | all 16K | 78.2 GB |
| H100 80 GB | 26 | all 16K | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | 35 | all 16K | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | 43 | all 16K | 107 GB |
| H200 141 GB | 61 | all 16K | 138 GB |
| B200 180 GB | 84 | all 16K | 176 GB |
| 2× RTX 4090 24 GB tensor parallel | 9 | all 16K | 23.4 GB a card |
| 2× RTX 3090 24 GB tensor parallel | 9 | all 16K | 23.4 GB a card |
| 2× RTX 5090 32 GB tensor parallel | 18 | all 16K | 31.0 GB a card |
| Requests at once | 8K tokens each |
|---|---|
| 1 | 31.8 GB |
| 5 | 38.5 GB |
| 8 | 43.5 GB |
| 16 | 56.9 GB |
| 32 | 83.8 GB |
| 64 | 137 GB |
One card, with vLLM's small-card settings.
Builds
| Build | Params | Precision | Weights | Smallest setup |
|---|---|---|---|---|
| phi-4 (above) ↗ | 14.7B | BF16 | 29.3 GB | L40S 48 GB |
| phi-4-FP8-dynamic ↗ packaged by RedHatAI |
14.7B | FP8 | 15.7 GB | 2× RTX 3060 12 GB |
| phi-4-GGUF ↗ packaged by unsloth 8 builds: Q2_K 5.6 GB … F16 29.3 GB |
— | GGUF | 8.9 GB | RTX 3060 12 GB |
From the model card
| Developers | Microsoft Research |
| Description | phi-4 is a state-of-the-art open model built upon a blend of synthetic datasets, data from filtered public domain websites, and acquired academic books and Q&A datasets. The goal of this approach was to ensure that small capable models were trained with data focused on high quality and advanced reasoning.phi-4 underwent a rigorous enhancement and alignment process, incorporating both supervised fine-tuning and direct preference optimization to ensure precise instruction adherence and robust safety measures |
| Architecture | 14B parameters, dense decoder-only Transformer model |
| Inputs | Text, best suited for prompts in the chat format |
| Context length | 16K tokens |
| GPUs | 1920 H100-80G |
| Training time | 21 days |
| Training data | 9.8T tokens |
| Outputs | Generated text in response to input |
| Dates | October 2024 – November 2024 |
| Status | Static model trained on an offline dataset with cutoff dates of June 2024 and earlier for publicly available data |
| Release date | December 12, 2024 |
| License | MIT |
| Primary Use Cases | Our model is designed to accelerate research on language models, for use as a building block for generative AI powered features. It provides uses for general purpose AI systems and applications (primarily in English) which require:1. Memory/compute constrained environments.2. Latency bound scenarios.3. Reasoning and logic. |
| Out-of-Scope Use Cases | Our models is not specifically designed or evaluated for all downstream purposes, thus:1. Developers should consider common limitations of language models as they select use cases, and evaluate and mitigate for accuracy, safety, and fairness before using within a specific downstream use case, particularly for high-risk scenarios.2. Developers should be aware of and adhere to applicable laws or regulations (including privacy, trade compliance laws, etc.) that are relevant to their use case, including the model’s focus on English.3. Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under. |
Our training data is an extension of the data used for Phi-3 and includes a wide variety of sources from:
Publicly available documents filtered rigorously for quality, selected high-quality educational data, and code.
Newly created synthetic, “textbook-like” data for the purpose of teaching math, coding, common sense reasoning, general knowledge of the world (science, daily activities, theory of mind, etc.).
Acquired academic books and Q&A datasets.
High quality chat format supervised data covering various topics to reflect human preferences on different aspects such as instruct-following, truthfulness, honesty and helpfulness.
Multilingual data constitutes about 8% of our overall data. We are focusing on the quality of data that could potentially improve the reasoning ability for the model, and we filter the publicly available documents to contain the correct level of knowledge.
We evaluated phi-4 using OpenAI’s SimpleEval and our own internal benchmarks to understand the model’s capabilities, more specifically:
MMLU: Popular aggregated dataset for multitask language understanding.
MATH: Challenging competition math problems.
GPQA: Complex, graduate-level science questions.
DROP: Complex comprehension and reasoning.
MGSM: Multi-lingual grade-school math.
HumanEval: Functional code generation.
SimpleQA: Factual responses.
phi-4 has adopted a robust safety post-training approach. This approach leverages a variety of both open-source and in-house generated synthetic datasets. The overall technique employed to do the safety alignment is a combination of SFT (Supervised Fine-Tuning) and iterative DPO (Direct Preference Optimization), including publicly available datasets focusing on helpfulness and harmlessness as well as various questions and answers targeted to multiple safety categories.
Prior to release, phi-4 followed a multi-faceted evaluation approach. Quantitative evaluation was conducted with multiple open-source safety benchmarks and in-house tools utilizing adversarial conversation simulation. For qual
Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.