THE ENGINE
Beryl LLM
is the pipeline
Every response Eve gives is routed through a purpose-built inference pipeline: Qwen2.5 for reasoning, GLM-4 for emotional depth, Hermes-3 for tool orchestration — all accessed via the HF Router with ZeroGPU burst capability. This is not a wrapper. This is Beryl LLM.
Qwen2.5-72B
Reasoning & Analysis
Qwen2.5-7B
Speed & Conversational
GLM-4
Emotional Intelligence
Hermes-3
Tool Use & Orchestration
ZeroGPU Local
On-device GPU Boost
CAPABILITIES
Everything You Need.
NOTHING YOU DON'T.
GET BERYL DESKTOP
Download & Install
One installer. No Python. No dependencies. Beryl LLM runs on HF infrastructure — your machine just needs a browser and internet.
⌘
macOS
Apple Silicon · M1/M2/M3
COMING SOON
🐧
Linux
AppImage · Ubuntu/Debian
COMING SOON
BERYL LIVE — DESKTOP COMING SOON
The full Beryl Live streaming avatar experience — Eve rendered in real-time via Runway AI — is coming natively to the desktop app. One click to go live.
NOTIFY ME ›
5+
LLM Models
6
Computer Use Tools
75MB
Installer Size
Free
ZeroGPU Inference
BEFORE YOU INSTALL
Try It LiveFULL BERYL LLM ON HUGGINGFACE
The exact same Beryl Desktop interface — Eve panel, chat, model tabs, ZeroGPU tiers — runs live on our HuggingFace Space. Open it in any browser, no install required.
Open Beryl on HuggingFace ›BERYL DESKTOP · POWERED BY BERYL LLM · FREE OPEN BETA
← Back to Beryl Live