Model family
Phi Models
Microsoft Phi models focus on small language models, edge deployment, reasoning efficiency, and low-resource local experiments.
Best for
Edge
Use this family hub to compare Phi variants for edge workflows, then open the detail page for deeper deployment notes.
Small
Use this family hub to compare Phi variants for small workflows, then open the detail page for deeper deployment notes.
Vision
Use this family hub to compare Phi variants for vision workflows, then open the detail page for deeper deployment notes.
Local
Use this family hub to compare Phi variants for local workflows, then open the detail page for deeper deployment notes.
Variants
Phi models grouped by workflow
Latest / flagship
Phi-4 Mini
Microsoft · Phi
Best for: Builders testing small local models on laptops, CPUs, and constrained hardware.
Local: A small-model option for local experiments on laptops, CPUs, and low-VRAM machines.
Phi-4
Microsoft · Phi
Best for: Small language model reasoning and assistant workflows
Local: Use the exact checkpoint and quantization that matches your hardware and latency target.
Phi-4 Multimodal
Microsoft · Phi
Best for: Small multimodal and edge experiments
Local: Use the exact checkpoint and quantization that matches your hardware and latency target.
Phi-3.5 MoE
Microsoft · Phi
Best for: Efficient reasoning model comparisons
Local: Use the exact checkpoint and quantization that matches your hardware and latency target.
Phi-3.5 Vision
Microsoft · Phi
Best for: Small vision-language experiments
Local: Use the exact checkpoint and quantization that matches your hardware and latency target.
Coding
Reasoning
Local-friendly
Phi-3 Mini
Microsoft · Phi
Best for: Local assistant workflows on 4–8 GB VRAM consumer GPUs, on-device inference, pipeline-embedded classification, and RAG retrieval on constrained hardware.
Local: Runs comfortably on a 6 GB or 8 GB consumer GPU at Q4_K_M. Use `ollama run phi3:mini` for the fastest local start.
Phi-3 Small
Microsoft · Phi
Best for: Small local assistant workflows
Local: Use the exact checkpoint and quantization that matches your hardware and latency target.
Phi-2
Microsoft · Phi
Best for: Legacy small model baseline
Local: Use the exact checkpoint and quantization that matches your hardware and latency target.
Compare
All Phi models in the directory
| Model | Type | Best for | Local runner notes | License | Detail |
|---|---|---|---|---|---|
| Phi-4 Mini | Edge | Builders testing small local models on laptops, CPUs, and constrained hardware. | A small-model option for local experiments on laptops, CPUs, and low-VRAM machines. | MIT | Open |
| Phi-4 | Reasoning | Small language model reasoning and assistant workflows | Use the exact checkpoint and quantization that matches your hardware and latency target. | Check exact model card | Open |
| Phi-4 Multimodal | Multimodal | Small multimodal and edge experiments | Use the exact checkpoint and quantization that matches your hardware and latency target. | Check exact model card | Open |
| Phi-3.5 MoE | Reasoning | Efficient reasoning model comparisons | Use the exact checkpoint and quantization that matches your hardware and latency target. | Check exact model card | Open |
| Phi-3.5 Vision | Vision | Small vision-language experiments | Use the exact checkpoint and quantization that matches your hardware and latency target. | Check exact model card | Open |
| Phi-3 Medium | Chat | Local assistant, reasoning, and long-context RAG workflows on 12–16 GB VRAM GPUs (RTX 3080/4070/4080). Good step up from 7B without requiring multi-GPU. | Use `ollama run phi3:medium` for the fastest local start. Q4 fits a 12 GB GPU; Q8 suits 16 GB or 24 GB setups. | MIT | Open |
| Phi-3 Mini | Edge | Local assistant workflows on 4–8 GB VRAM consumer GPUs, on-device inference, pipeline-embedded classification, and RAG retrieval on constrained hardware. | Runs comfortably on a 6 GB or 8 GB consumer GPU at Q4_K_M. Use `ollama run phi3:mini` for the fastest local start. | MIT | Open |
| Phi-3 Small | Edge | Small local assistant workflows | Use the exact checkpoint and quantization that matches your hardware and latency target. | Check exact model card | Open |
| Phi-2 | Edge | Legacy small model baseline | Use the exact checkpoint and quantization that matches your hardware and latency target. | Check exact model card | Open |
| Phi-3 Medium 128K Instruct | Chat | Phi-3 Medium 128K Instruct is Microsoft's larger Phi-3 checkpoint, pairing 14B parameters with the same 128K context window as the smaller Phi-3 Mini. | 14B with 128K context. Strong reasoning for size; context-heavy RAG tasks benefit from the long window. Q4_K_M weights measured at 8.57 GB, so 12 GB is the practical minimum — on an RTX 3080 (10 GB) 9% of resident bytes spilled to host (measured 2026-08-07, Ollama 0.32.5). Note the tag: the friendly `phi3:medium` serves Q4_0 (7.9 GB), not the Q4_K_M these figures describe, so this entry points at the explicit Q4_K_M tag instead. Throughput is unmeasured here because this model stops generating early against the benchmark's fixed prompt (163 tokens when 512 were requested) — the same behaviour Phi-3 Mini shows at 107 tokens, so it looks architectural rather than model-specific. | MIT | Open |