Model
Model explorer

Phi-4-multimodal (5.6B)

OPEN
Microsoft · Phi-4 family · released Feb 26, 2025

5.6B unified text+vision+speech model on frozen Phi-4-Mini weights with LoRA adapters; ranked #1 on the HF OpenASR leaderboard (6.14% WER) at release.

ReasoningCodingVisionFunction callingTool useAgentic
842.4
Elo · unrated
Parameters
5.6B
Active params
5.6B (dense)
Context
128K tokens
Architecture
Dense transformer (Phi-4-Mini backbone) + mixture-of-LoRAs vision/speech adapters (5.6B)
License
MIT
Languages
23+
API price (in/out)
$0.08 / $0.32
Modalities
text · vision · audio
Benchmark results
Bar shows position within the tracked field; marker = field best
AI2DVision82.3%#31
best: Molmo 72B · 96.3%
ChartQAVision81.4%#30
best: MiniMax-VL-01 · 91.7%
DocVQAVision93.2%#21
best: Qwen2-VL-72B · 96.5%
MathVistaVision62.4%#40
best: Seed 2.1 Pro · 90.7%
MMBench (English)Vision86.7%#11
best: Qwen3.5-397B-A17B · 93.7%
MMMU-ProVision38.5%#48
best: Claude Opus 4.7 · 85.5%
MMMUVision55.1%#85
best: Claude Fable 5 · 89.3%
OCRBenchVision84#20
best: InternVL3-78B · 906
TextVQAVision75.6%#17
best: Molmo 2 8B · 85.7%
Video-MMEVision55.0%#22
best: Seed 2.1 Pro · 89.2%
Run it locally
VRAM @ Q4
4 GB
VRAM @ FP16
12 GB
Fits on (Q4)
RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB
No independently verified RTX 4090 llama.cpp benchmark found; the 5.6B size runs comfortably above real-time on a single 4090.
Quantizations
GGUF Q4 · ONNX INT4
Phi-4 family
Elo progression across releases
Phi-4 (14B)Dec 20241016.7Phi-4-mini (3.8B)Feb 2025529.8Phi-4-mini-instructFeb 2025523.2Phi-4-multimodal (5.6B)Feb 2025842.4Phi-4-multimodal-instructFeb 2025842.4Phi-4-reasoningApr 20251292.1Phi-4-reasoning (14B)Apr 20251292.1
Resources
HFmicrosoft/Phi-4-multimodal-instructDOCSazure.microsoft.com/en-us/blog/empowering-innovation-the-next-generation-of-the-phi-familyAPIAvailable via Microsoft API
API price $0.08/$0.32 · each benchmark row carries its own source badge (see methodology)