Migrate to Responses
Use the OpenAI Python SDK to callhttps://api.parallel.ai/v1/responses. Pointing an existing Chat Completions client or ChatParallel instance at a new model name is not sufficient. Responses uses a different endpoint and request shape.
Install the SDK:
PARALLEL_API_KEY to your Parallel API key, then create a response:
model="parallel" and select a research tier with reasoning.effort: low, medium, high, or auto. The default effort is medium. Higher effort performs more research; high requests can take up to a minute.
The Responses API uses input for the request. Read the answer through the SDK’s response.output_text accessor. Do not reuse Chat API model names or assume that its response metadata has the same shape.
Use Parallel with LangChain
For agent tool calls, use the Parallel search tool and extract tool with a tool-calling model. These integrations use the separate Search and Extract APIs. For a research tool backed by Responses, see Parallel’s research subagent guide.See also
Connect these docs to your agent of choice via MCP for real-time answers.

