New Seven agents, one intelligence layer · Hindi + English
model / claude-sonnet-4-6
Sonnet 4.6 Opus 4.6 Haiku 4.5
The intelligence layerpowering live commerce. RunAsh is built on Claude — tuned for real-time stream analysis, agentic commerce workflows, and sub-second CTA decisions. Paired with an open research stack spanning vision, speech, and generative video.
200K tokens Context
~70 tok / sec Speed
<1s TTFT Latency
4,096 tokens Max output
Overview Playground Capabilities Benchmarks Research Changelog
chat
Ask RunAsh anything. RunAsh Chat-Agent
secure proxy · /api/chat
Ask anything about live streams, CTAs, scripts or commerce strategy.
Write 3 urgency CTAs for a limited sneaker drop How should I pace a 30-minute live shopping stream? Turn 'Only 3 left in stock!' into a Hindi + English hype line
Enter to send · Shift+Enter for a new line live capability demos
Try it. Right now. Pick one to load it into the composer above CTA Generator
Product + price, e.g. 'Wireless earbuds, $49'
Intent Classifier
Paste viewer chat messages, one per line
Script Writer
Product + duration, e.g. 'AirPods Max, 45 seconds'
Translator
Text to translate, e.g. 'Only 3 left in stock!'
Sentiment Pulse
Paste a stream transcript or viewer chat batch
Agentic Planner
Commerce goal, e.g. 'Sell 100 units in 1 hour'
inference pipeline
Stream in. Action out. 01
Stream ingestion RTMP / WebRTC
02
Scene parsing Vision + ASR
03
Intent model Commerce signals
04
RunAsh LLM claude-sonnet-4-6
05
Action dispatch CTA · Overlay What the model handles Live CTA timing and copy generation Real-time product description rewriting Viewer sentiment classification from chat Inventory-aware upsell recommendations Multi-platform caption localisation Post-stream analytics narrative Seller coaching and script feedback Payment link generation in-stream Ready to go live? Start your first AI-powered stream in under 5 minutes.
