Chat Dashboard Performance Demo Games

mμ|LLM

Local-first AI router. Microscopic costs.
CLASSIFY → ROUTE → CHEAPEST CAPABLE MODEL
LIVE STATS
Queries
—
Savings
—
Cache Rate
—
Local Rate
—
HOW ROUTING WORKS

Every query takes the cheapest path

muLLM classifies your intent, estimates complexity, and routes to the cheapest model that can handle it. Simple queries? Free from cache or your local GPU. Complex? Cloud, but only when truly needed.
CACHE
free, instant
→
LOCAL GPU
free, your RTX
→
CLOUD
paid, powerful
FEATURES
⚡

Vector Cache

ChromaDB-backed semantic cache. Similar questions hit cache instantly. Saves money on repeated queries.

—
cached responses
💻

Local-First

qwen3.5:9b on your RTX 5090. Handles 80%+ of queries for free. No data leaves your machine.

—
queries handled locally
☁

Smart Escalation

Complex queries auto-escalate to cloud (Anthropic, OpenAI, Google). 12 models configured, auto-picks cheapest.

—
cost per hour
🔀

Split Routing

Multi-part prompts decomposed and routed in parallel. Each sub-task gets the cheapest capable model.

🎤

Voice I/O

Mic input + speaker output. On-device speech recognition and synthesis. Works offline, PWA compatible.

⚙

Self-Orchestration

muLLM can update itself. The orchestrator reads task files, generates patches, and applies them via code-gen mode.

TRY IT

See routing in action

Click a sample query to see how muLLM routes it. Watch the routing animation above.
mu|LLM Chat — Demo Mode
COST COMPARISON

Real savings, real numbers

These are actual savings from this muLLM instance, not estimates.

vs Claude Sonnet

—
saved

vs Claude Opus Extended

—
saved

vs Blended Average

—
saved
Start Chatting →
Dashboard Performance