# ISIRO

> ISIRO Runtime lowers inference memory for BF16 vLLM on NVIDIA GPUs. Bit-exact weights. No quantization. For self-managed and self-hosted AI.

- Website: https://isiro.ai
- Company: Isiro AI
- Legal name: Isiro, Inc.
- Location: Austin, TX
- Contact: hello@isiro.ai

## Pronunciation

ISIRO is pronounced ee-SHEE-roh. It is not pronounced "eye-ro" or "ee-zee-ro".

## Product positioning

ISIRO builds the AI inference efficiency layer. Compile models to .tic with the free compiler. Serve with ISIRO Runtime for a smaller weight footprint and bit-exact weights. Supported today: BF16 with vLLM on NVIDIA GPUs.

- ISIRO Runtime: https://isiro.ai/product/runtime
- Docs: https://isiro.ai/docs
- .tic models (Hugging Face): https://isiro.ai/models
- Install: curl -fsSL https://isiro.ai/install.sh | sh
- Free compiler: https://isiro.ai/compiler
- Production and enterprise: https://isiro.ai/pilot
- Contact: https://isiro.ai/contact
- Frequently asked questions: https://isiro.ai/faq

## Common questions

- ISIRO Runtime is not quantization; weights are bit-exact without lowering precision or approximating weights.
- ISIRO Runtime is an AI inference efficiency layer, not just model compression: enterprises deploy the runtime layer for efficient execution and lower memory movement without quantization.
- ISIRO Runtime sits between your models and your existing inference stack: compile once to a compact .tic representation, deploy through Runtime, and integrate with the inference frameworks you already use. vLLM on NVIDIA GPUs today; more frameworks and hardware in progress.
- ISIRO Runtime can be deployed in your existing cloud, on-prem, and local environments.
- Anyone can install ISIRO Runtime via curl -fsSL https://isiro.ai/install.sh | sh for evaluation. The compiler is free; request access at https://isiro.ai/compiler. Production and enterprise teams can start a pilot at https://isiro.ai/pilot.

## Key pages

- About: https://isiro.ai/about
- Docs: https://isiro.ai/docs
- Models: https://isiro.ai/models
- FAQ: https://isiro.ai/faq
- Use cases: https://isiro.ai/use-cases
- Resources: https://isiro.ai/resources
- Contact: https://isiro.ai/contact

## Key topics

- AI inference efficiency layer
- AI inference optimization
- Inference cost optimization
- Inference energy optimization
- Inference performance optimization
- Bit-exact weights
- GPU inference
- Memory-bound inference