The whole model, in your chat app.
Sovereign AI lives on Telegram. Ask it anything crypto. 87% A-rate on our benchmark. Start free.
Commands
/ask <question>/analyze <token>/predict <ticker> [tf]/sentiment <text>/pricing/tier/upgrade/feedback <text>Pricing
- •Unlimited queries
- •30 req/min rate limit
- •Priority on busy days
- •Future: $SVRN staking unlocks Pro (500/mo)
How a query works
- 1. Your message goes to the bot.
- 2. The bot checks your tier and daily usage in Redis.
- 3. If allowed, it calls
sovereign-apiwhich runs the query on our V100 Ollama instance. - 4. Response comes back with a non-removable educational disclaimer.
- 5. Typical latency: 2–8 seconds.
FAQ
Is this financial advice?+
No. Every paid response carries a non-removable educational disclaimer. We discuss mechanics and scenarios; we do not call entries or exits.
What model is this?+
A fine-tune of Qwen3-14B, trained on crypto-specific instruction data using our in-house SFT + DPO pipeline. Details and eval methodology are on the research page.
Where does it run?+
On our own hardware — a Tesla V100 16GB with an RTX 2070 as a secondary worker. Not a wrapper on OpenAI.
What if the model is wrong?+
It will be wrong sometimes — all LLMs are. We publish the full benchmark methodology so you can see where it fails. Cross-check anything before acting.
Can I use this for automated trading?+
The free and Pro tiers are for manual, educational use. For programmatic access, use the REST API with enterprise-tier rate limits. See /api.
Does paying require KYC?+
Crypto payment (SOL): no KYC. Fiat (Stripe): standard card processing, no KYC.