Kumari

Gold Verified

Publisher: EverestQ Engineering · Category: Chat

Parameters

Context Window

32K

Default Size

5.6 GB

Downloads

289,100

Ultra-fast low-latency assistant model optimized for real-time conversational streaming and sub-10ms response times.

PULL MODEL

eq pull everestq/kumari

RUN INFERENCE

eq run everestq/kumari "Summarize quantum computing principles"

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.