Blog

Blog

Twelve pages, each for one job: pick a model, point an IDE, or start in Telegram.

Models

• What GLM 5.3 Flash Uncensored isDefault model: chat, text, code, fewer refusal theatrics.
• Why the model writes a refusalRefusals are a pattern in the weights. Why the catalog has six models.
• Which model to chooseSix ids without cloning the price list: GLM, Qwen, Abliterated.

API and IDEs

• Your Base URL instead of api.openai.comSame SDK, different host. Chat Completions only.
• How to point Cursor at this APICursor only: Base URL /v1, bot key, new chat.
• Why the API returns 401, 402, or 404Status codes: key, balance, wrong protocol.
• Why PAYG if you can rent your own boxTokens instead of an hourly GPU rental.

Start and Telegram

• Path: Telegram → key → APIThree steps: bot, key, Base URL. No web cabinet.
• Chat with the model in TelegramBot DM is chat. API is a key plus /v1.
• How to issue an API keyKeys only in the bot, shown once, no site cabinet.
• How a request is pricedTokens, $0.005 minimum, packs in the bot. Not a clone of the price cards.
• Research and responsibilityResearch-only without legal theater. The request is yours.