Tencent

Hy3

hy3

Tencent's 295B-parameter mixture-of-experts model for reasoning and agentic workflows, with 21B active across 192 experts and top-8 routing. Reasoning is on by default at an effort level you set per request, and can be turned off entirely.

Input: Text
Context window: 262K tokens
Precision: 4 bits
Price per 1M tokens: Coming soon
Integration: OpenAI-compatible API

ReasoningConfigurable effortMoEAgentic coding

Playground

Hy3

Preview unavailable

Explore this model in the app

The inline preview isn’t available right now. Check the chat app for availability.

Open chat ↗

Limited free previewContinue in chat ↗

Use it in your app

API documentation ↗

Client Libraries

curl https://api.elastix.ai/v1/chat/completions \
  -H "Authorization: Bearer $ELASTIX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "hy3",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'