AI Telegram Bot in 2026: python-telegram-bot (29k Stars) vs aiogram vs grammY - Build an LLM Bot in 15 Minutes

Telegram has the friendliest bot API of any chat app. python-telegram-bot (29,383 stars), aiogram (5,818) and grammY (3,705) all connect to it - here is how to pick and build an LLM bot fast.

💡 What You Will Learn

Telegram has the friendliest bot API of any chat app. python-telegram-bot (29,383 stars), aiogram (5,818) and grammY (3,705) all connect to it - here is how to pick and build an LLM bot fast.

## The short answer **python-telegram-bot** (29,383 stars, GPL-3.0) is the most battle-tested Python wrapper - great docs, huge community, sync or async. **aiogram** (5,818 stars, MIT) is the modern async choice: faster, cleaner code, FSM support for multi-step dialogs. **grammY** (3,705 stars, MIT) is the TypeScript framework with a plugin ecosystem. ## 15-minute bot with python-telegram-bot ```bash pip install python-telegram-bot openai ``` ```python from telegram import Update from telegram.ext import Application, CommandHandler, MessageHandler, filters import openai TOKEN = "YOUR_BOT_TOKEN" async def chat(update: Update, context): r = openai.chat.completions.create( model="gpt-4o-mini", messages=[{"role": "user", "content": update.message.text}]) await update.message.reply_text(r.choices[0].message.content) app = Application.builder().token(TOKEN).build() app.add_handler(MessageHandler(filters.TEXT, chat)) app.run_polling() ``` Get the token from @BotFather: `/newbot` - name - token. ## Picking the right framework | Framework | Stars | Language | Best for | |:----------|------:|:---------|:---------| | python-telegram-bot | 29,383 | Python | Docs, stability, sync/async | | aiogram | 5,818 | Python | Async, multi-step dialogs (FSM) | | grammY | 3,705 | TypeScript | Plugins, session middleware | ## Real numbers - `run_polling()` handles thousands of users on one process - Telegram's API is pull-based, no webhooks needed for small bots. - A bot replying via LLM API costs roughly $0.0002-0.004 per message with gpt-4o-mini-class models, or $0 with local Ollama. - python-telegram-bot receives about 1 million downloads per month. ## FAQ **Q: Webhook or polling?** A: Polling is fine up to ~10k users. Above that, use webhooks with a domain and HTTPS. **Q: Can the bot remember conversations?** A: Yes - store message history per user in a dict, Redis, or SQLite, and send the last N messages as context.