// Open-model VRAM table · family
Phi: local deployment & VRAM
18 open-weight models from Microsoft (28 official repos incl. quantised builds). At 4-bit they range from 3.6 GB (Phi-tiny-MoE-instruct) to 29.5 GB (Phi-3.5-MoE-instruct).
official orgs: microsoftlatest 2026-01-23memory = estimate
01Text LLM (14)
| Model | Params | Context | BF16 | 4-bit | Smallest GPU (4-bit) | Mac (4-bit) | Licence | Released |
|---|---|---|---|---|---|---|---|---|
| Phi-mini-MoE-instruct | 7.6B / 2.4B | 4K | 18.3 GB | 6.6 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-06-23 |
| Phi-tiny-MoE-instruct | 3.8B / 1.1B | 4K | 9.4 GB | 3.6 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-06-23 |
| Phi-4-mini-flash-reasoning ↗ | 3.85B | 256K | 10.3 GB | 4.4 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-06-19 |
| Phi-4-mini-reasoning | 3.84B | 128K | 10.0 GB | 4.1 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-04-29 |
| Phi-4-reasoning-plus | 14.7B | 32K | 34.4 GB | 12.0 GB | 16GB GPU | Mac 24GB | Commercial OK | 2025-04-17 |
| Phi-4-reasoning | 14.7B | 32K | 34.4 GB | 12.0 GB | 16GB GPU | Mac 24GB | Commercial OK | 2025-04-09 |
| Phi-4-mini-instruct | 3.84B | 128K | 10.0 GB | 4.1 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-02-19 |
| phi-4 | 14.7B | 16K | 34.4 GB | 12.0 GB | 16GB GPU | Mac 24GB | Commercial OK | 2024-12-11 |
| Phi-3.5-MoE-instruct | 41.9B / 6.6B | 128K | 93.8 GB | 29.5 GB | 2 × RTX 5090 (64GB) | Mac 64GB | Commercial OK | 2024-08-17 |
| Phi-3.5-mini-instruct | 3.82B | 128K | 12.1 GB | 6.3 GB | 8GB GPU | Mac 16GB | Commercial OK | 2024-08-16 |
| Phi-3-medium-128k-instruct | 14B | 128K | 32.9 GB | 11.5 GB | 16GB GPU | Mac 24GB | Commercial OK | 2024-05-07 |
| Phi-3-medium-4k-instruct | 14B | 4K | 32.9 GB | 11.5 GB | 16GB GPU | Mac 24GB | Commercial OK | 2024-05-07 |
| Phi-3-small-128k-instruct ↗ | 7.39B | 128K | 17.8 GB | 6.5 GB | 8GB GPU | Mac 16GB | Commercial OK | 2024-05-07 |
| Phi-3-small-8k-instruct | 7.39B | 8K | 17.8 GB | 6.5 GB | 8GB GPU | Mac 16GB | Commercial OK | 2024-05-07 |
02Vision-language (3)
| Model | Params | Context | BF16 | 4-bit | Smallest GPU (4-bit) | Mac (4-bit) | Licence | Released |
|---|---|---|---|---|---|---|---|---|
| Phi-4-reasoning-vision-15B | 15.1B | 32K | 35.4 GB | 12.3 GB | 16GB GPU | Mac 24GB | Commercial OK | 2026-01-23 |
| Phi-3.5-vision-instruct | 4.15B | 128K | 12.8 GB | 6.5 GB | 8GB GPU | Mac 16GB | Commercial OK | 2024-08-16 |
| Phi-3-vision-128k-instruct | 4.15B | 128K | 12.8 GB | 6.5 GB | 8GB GPU | Mac 16GB | Commercial OK | 2024-05-19 |
03Speech-to-text (1)
| Model | Params | BF16 | 4-bit | Smallest GPU (4-bit) | Mac (4-bit) | Licence | Released |
|---|---|---|---|---|---|---|---|
| Phi-4-multimodal-instruct | 5.57B | 12.8 GB | 4.2 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-02-24 |
04Other families
Qwen · IBM Granite · NVIDIA Nemotron · Gemma · Mistral / Magistral / Devstral · GLM / CogVideoX (Z.ai) · DeepSeek · Ai2 OLMo / Molmo · Ant Ling / Ring / Ming · Liquid LFM · MiniCPM / VoxCPM · Tencent Hunyuan / Hy · Llama · Meituan LongCat · Jina embeddings & rerankers · Wan · Kimi · Cohere Command / Aya · Xiaomi MiMo · FLUX · Microsoft VibeVoice / Fara / Harrier · InternLM / Intern-S · StepFun Step · MiniMax · Baidu ERNIE · NVIDIA Parakeet / Canary speech · Stable Diffusion / Stable Audio · Yi (01.AI) · ByteDance Seed · SmolLM / SmolVLM · BGE embeddings & rerankers · LTX Video · FunAudioLLM speech · gpt-oss · poolside Laguna · Whisper · Baichuan · Thinking Machines Inkling · xAI Grok · rednote dots
Totals include 10% + 0.5GB overhead and one 8K-token KV cache for text models. Estimates only. Method →