Model Families
Privacy-first foundation models built for on-device inference. Pure Rust, zero external ML dependencies, cross-platform deployment.
SLM360 Nano
Privacy-First Encoder for On-Device Understanding
A 6.4M-parameter bidirectional encoder optimized for intent classification and NLU. Achieves sub-5ms latency with INT4 quantization, 100% offline, zero external ML dependencies. Built in pure Rust with SIMD acceleration.
6.4M
Parameters
4MB
INT4 Size
<5ms
Latency
94.1%
Accuracy
SLM360 Base
On-Device Decoder for Text Generation
A 125M-parameter causal decoder for autoregressive text generation. Sub-50ms per token with INT4 quantization. Weight-tied LM head, Grouped Query Attention with 3x KV cache reduction, and LoRA adapters for per-user personalization.
125M
Parameters
63MB
INT4 Size
<50ms/tok
Latency
2,048
Context
Med360 Lite
Medical AI Assistant for Indian Healthcare
Fine-tuned on Google Gemma 3 4B with 173K+ training examples. Native Hinglish support, Indian pharmaceutical nomenclature, and AIIMS/NEET-PG level medical knowledge. Optimized for mobile and edge deployment.
4B
Parameters
173K+
Training Data
Gemma 3
Base Model
Hinglish
Languages
3 model families