DeepSeek Flash ⚡ — self-hosted on ZeroGPU

The model deepseek-ai/DeepSeek-R1-0528-Qwen3-8B runs inside this Space on ZeroGPU — no external inference provider. It reasons internally and streams the final answer.

API usage (a token charges your ZeroGPU quota; sign in for more):

from gradio_client import Client
c = Client("Desmondo93/deepseek-flash-demo", oauth_token="hf_...")
print(c.predict("hi", api_name="/respond"))

Agents: connect an MCP client to https://desmondo93-deepseek-flash-demo.hf.space/gradio_api/mcp/.