Gemma 4 31B IT
gemma4-31b-itGoogle's multilingual model for conversation, reasoning, and coding.
MultilingualReasoningCoding
Talk to open models in Uzbek, Russian and English, or connect your application through the CoreAI API. Both run on our GPUs in Tashkent.
Talk to a curated set of open models in Uzbek, Russian, or English. Streaming replies, code, copy, and history.
Use JSON requests or streamed events from your backend. API keys are available after registration.
https://inference-api.coreai.uz/v1Two models are live in chat and through the API. One more is on the roadmap.
gemma4-31b-itGoogle's multilingual model for conversation, reasoning, and coding.
qwen3.8-27bA dense model with a 262K-token context window for reasoning, coding, and tool-driven workloads.
deepseek-ai/DeepSeek-V4-Flash-0731A 304B open-weight model for advanced coding, reasoning, and long-running agent workflows.
Our own GPUs in a Tier III datacenter in Tashkent. Your data stays in the country, and the capacity is ours to run.
Chat works in Uzbek, Russian and English, including Uzbek Latin and Cyrillic.
A separate environment for banks and government, handled under local data law (ZRU-547).
Chat is free during the public testing period. You don't need an account to start.
Create a Chat/API account, generate a key in the API console, and follow the live documentation.
On our GPUs in a Tier III datacenter in Tashkent. Your data does not leave Uzbekistan.
Free-tier conversations may be reviewed to evaluate quality, understand usage, and improve CoreAI. See the privacy policy for details.
Run on the same GPUs in our cloud, or deploy a private cloud of your own.