LLM chat completion: send messages and max_tokens, get the reply. OpenAI-style chat completion from Qwen3-32B. Returns: an OpenAI-style chat completion from the model the service is named for (DeepSeek-V4-Flash-0731, Qwen3-32B or Llama-3.1-8B-Instruct): POST {messages: [{role, content}], max_tokens <= 1024}; returns the assistant message, the model, and its token usage. Messages up to 16,000 characters in total.