LLM
GA
Open weights
Phi-4 Multimodal
Visual + audio understanding on edge devices
Release2025-02
Context131K
Max output—
Parameters5.6B
Local / VRAM8–16 GB
LicenseOpen weights
Overview
Phi-4 Multimodal is a model in the LLM category from Microsoft. It was released in 2025-02. The vendor is based in US. Best for: Visual + audio understanding on edge devices.
Strengths
- Very small
- Audio/visual
- MIT
Limits
- Weak reasoning
- Short outputs
Specifications
| Vendor | Microsoft |
|---|---|
| Category | LLM |
| Release | 2025-02 |
| Status | GA |
| License | Open weights |
| License name | MIT |
| Commercial use | Allowed |
| Parameters | 5.6B |
| Architecture | VLM + audio encoder |
| Context window | 131K |
| Input modalities | Text, Image, Audio |
| Output modalities | Text |
| Prompt types | Text |
| Negative prompts | No |
| Max outputs | 1 |
| Local deployment | Yes |
| VRAM | 8–16 GB |
| System RAM | 16–32 GB |
| API access | Yes |
| Pricing | Free weights |
| Confidence | High |
| Vendor country | US |
| Vendor type | Cloud provider |
Pricing & hardware
Free weights
This model can be self-hosted or run locally. Approximate GPU memory: 8–16 GB. Approximate system RAM: 16–32 GB. API access is available.