Questions and answers
Dikaya answers about the demo, subscriptions, API access, images, model reasoning and conversation storage. Public facts available without an account.
What is Dikaya?
Dikaya is an uncensored AI service for developers and writers, offering browser chat, IDE connections and an API with one prepaid balance. It provides three models, stores browser conversation history and shows usage costs. There is no subscription and balances do not expire.
What does “uncensored AI” mean?
Dikaya positions itself as uncensored AI for code, writing and creative tasks. This describes the service approach, not a guarantee that every request will receive an answer or that answers will be correct. Results depend on the model, prompt and available context.
Is there a free demo?
New eligible accounts can receive $0.5 and up to 25 Pro 2 requests. The demo is text-only, with 512 output tokens, 6000 bytes of context and 10 requests per minute. The bonus is issued once, subject to browser, network and overall limits. Previously granted conditions are preserved.
Do I need a subscription?
No. Dikaya uses prepaid credit without monthly charges or balance expiry. Top-ups start at $5; browser chat and API requests use the same balance.
What is the Dikaya API?
The Dikaya API connects the service models to applications and IDEs using a personal key. The base URL is https://dikaya.cloud/v1. It supports Chat Completions, Responses and Messages request formats, plus SSE streaming. Protocol compatibility does not imply support for every feature of a third-party client.
How are images handled?
Flash processes images directly. For paid Pro and Pro 2 requests, Flash first creates a text description for the selected model. Both stages are billed and reserved before processing begins. Descriptions can lose visual details. The demo remains text-only.
How do top-up bonuses work?
Bonuses depend on one confirmed payment: 5% from $20, 10% from $50 and 15% from $100. Payments are not combined to reach a higher tier. Credit is added after payment confirmation, not when an invoice is created.
Are reasoning tokens billed?
Output usage includes reasoning tokens when the model uses them. They are part of output cost, not a second additional charge. On Pro 2, reasoning_effort=none does not guarantee zero reasoning tokens.
What happens if a request is interrupted?
A conservative cost estimate is reserved before inference. An explicit HTTP rejection refunds unused work; completed image processing is billed separately. If the outcome or usage is unknown, the reservation is retained for reconciliation. Closing the browser tab does not cancel accounting.
Are my messages stored?
Browser chat history is stored in your account. Prompts, responses and request metadata are retained for manual moderation, accessible to the administrator in a protected console and private Telegram bot. Logs are not public.
How is the acquisition source recorded?
A tagged link or available referrer may set a signed Dikaya cookie for 30 days. Registration keeps the first tagged source in that window; confirmed payments can then be grouped by it. This analytics feature retains only a referrer hostname, not its full address, URL parameters or IP address. Older accounts are not assigned a source retroactively.
Is Dikaya Pro 2 GLM-5.3?
Dikaya Pro 2 is a modified version based on GLM-5.3, offered as uncensored AI, rather than the unmodified original model. It supports up to 1 million context tokens; demo and maximum-output limits are specified separately. Use seller-pro-2 in the API.
Practical guides
Guides on code, writing, GLM-5.3 and API integration (Russian)
Source: official Dikaya model catalogue, published rates and account/demo conditions. Pricing · Documentation
Facts reviewed: . Publisher: Dikaya.