Stand up Kimi K3
on Azure.

export MODEL_ID="moonshotai/Kimi-K3"
export INITIAL_CONTEXT="32768"
export MIN_TOTAL_VRAM_GB="3072"

# This model exceeds a single 8-GPU node.
# Use the linked official model recipe with the provider's cluster path.
# Keep the initial context conservative until a short smoke test passes.

This is a reproducible starting block. Cluster paths still need the model author's distributed recipe and the provider's orchestration layer.

curl http://127.0.0.1:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k3",
    "messages": [{"role": "user", "content": "Reply with: deployment healthy"}],
    "max_tokens": 32
  }'

Run this only after logs show the model loaded. A successful HTTP response is not a latency or quality benchmark.