Home › AI tools › AI chat and an LLM API, billed in Birr.
AI chat and an LLM API, billed in Birr.
Use Yakal Chat in the console or connect a compatible app to the API. The same wallet pays for chat tokens, so you can move from a browser conversation to a coding agent without opening a separate international account.
Create an account · Open on Telegram
Models for chat and agents
The catalog includes glm-5.3-flash, deepseek-v4.1-flash, glm-5.2 and glm-5.3. Faster flash models cost less per token; larger models consume quota units faster. Some models support long context, vision or tool calls, but capabilities vary by model.
- Use flash models for drafting, translation, summarisation and routine agent loops. Use larger models when the task needs deeper reasoning. Always check the model card and price before relying on a capability.
Languages and limits
Chat prompts support Amharic, Afaan Oromoo, Tigrinya and English. Output quality depends on the model and prompt. AI can make factual mistakes, so verify legal, medical, financial and other high-stakes answers with a qualified source.
- Long conversations consume more tokens because the model re-reads the conversation. Cached input can reduce the cost on supported models, but you should still watch the usage panel. Paid chat has a minimum charge per request.
OpenAI-compatible API
The base URL is https://api.yakal.et/v1. Create a Yakal API key, then point a compatible SDK or client at that base URL. Supported endpoints and model names are listed in the API documentation. OpenAI-compatible means the request shape is familiar; Yakal is a separate provider and is not endorsed by OpenAI.
- Per-key budgets and rate limits help contain a leaked key. Usage analytics show what each key spends. Test with a small request before sending production traffic.
Coding agents
The documentation includes setup examples for OpenCode, Pi, Codex and Hermes. You can use a Yakal key with an OpenAI-compatible base URL. Agent work can consume many tokens quickly because each turn resends context, so set a budget and choose a model deliberately.
Pricing
glm-5.3-flash starts at 90 ETB per million output tokens and 27 ETB per million input tokens. Other models have different rates, shown on the pricing and model pages. Free allocations apply to text chat only, not to media products.
Questions
Can I chat with AI in Amharic?
Yes, Yakal Chat accepts Amharic prompts. Quality and output language depend on the selected model. Review important answers, especially for legal, medical or financial topics.
Is there an OpenAI-compatible API in Ethiopia?
Yes. The base URL is https://api.yakal.et/v1 for supported endpoints and models. Use a Yakal API key and the documentation. It is a separate service, not OpenAI.
Which model should I use for coding agents?
Flash models are usually the economical choice for routine agent loops; larger models may help with harder reasoning. Check context limits, tool-call support and token prices for the model you select.
Are free tokens available for the API?
Eligible accounts may receive a one-time 5,000,000-token signup allocation and separate monthly Flash allowances. The signup tokens are usable 1:1 on any supported text model and apply to text chat. Media is wallet-billed.
Related guides
- Ethiopian AI, built for local languages and Birr payments.
- AI in Birr: pay for what you use, with Telebirr.
- AI images in Ethiopia: text to image, editing and design.
- AI video in Ethiopia: text to video, image to video and more.
- An AI website builder for Ethiopian businesses.
First-party product information, updated 2026-11-14. Prices, availability and eligibility can change. Check the live quote and Wallet. AI output can be wrong; verify important results.