Phi-4
Partner VerifiedPublisher: Microsoft · Category: Reasoning
Parameters
14B
Context Window
16K
Default Size
9.1 GB
Downloads
298,400
Microsoft state-of-the-art 14B parameter reasoning model trained with synthetic data curation techniques.
PULL MODEL
eq pull phi4
RUN INFERENCE
eq run phi4 "Explain vector database indexing algorithms"
Attention Architecture
Utilizes Grouped-Query Attention (GQA) with 16-token fixed Paged KV Cache block allocation for zero-fragmentation memory residency.
ONNX & EQC Compatible
Compiles directly through the EQC toolchain into standalone .eqx binary packages with fused SwiGLU kernels.
Multi-Hardware Support
Auto-detects CUDA RTX/A100/H100, Apple Silicon Metal Performance Shaders, or AVX-512 CPU execution backends.