Skip to content

Dialogue and Reasoning

This page explains how to use Grok series dialogue calls and reasoning scenarios. For an integration overview, see Grok Overview.

Basic Conversation

resp = client.chat.completions.create(
    model="grok-4.5",
    messages=[
        {"role": "system", "content": "You are a sharp but friendly tech commentator"},
        {"role": "user", "content": "Give an assessment of the recent AI industry trends"},
    ],
)
print(resp.choices[0].message.content)

Multi-turn conversations are the same as other series: include historical messages in messages, see Multi-turn Conversations;
For streaming output, add "stream": true, see Streaming Output.

Reasoning Scenarios

Grok high-tier models perform strongly on math, logic, and coding tasks:

For reasoning tasks, it is recommended to set max_tokens generously to leave room for the model to reason;
If the output includes a thinking process field (reasoning_content), display it separately from the main text, handle it the same way as Reasoning Model Output;
For complex tasks, first get the workflow working on a free/low-cost model, then switch to Grok to compare results.

Selection Recommendations

Social-media-style conversations, trending topic commentary: Grok primary tier;
Serious long-form content and compliance-sensitive scenarios: recommend comparing with the Claude series before choosing;
Use the model name configuration option for gradual rollout, comparing answer quality and cost for the same prompt.