Skip to main content
POST
Chat completion: Gemini 3.6 Flash (OpenAI-compatible)
Use the playground on the right: put Bearer sk-your-api-key in Authorization. The default example uses reasoning_effort: "low"; hit send to see a response.
Thinking is ON by default (bills as output). Keep it at low for debugging; switch to high with max_tokens at 4000+ for complex tasks. Search grounding, code execution, and other native tools are not available on this endpoint — use the Native Playground instead. Overview and measured data: Gemini 3.6 Flash Overview.

Parameter quick reference

Response notes

  • Thinking spend is in usage.completion_tokens_details.reasoning_tokens (reported for this model)
  • Thinking text (reasoning_content) is not echoed on this endpoint — use the native endpoint’s includeThoughts to inspect thoughts
  • Double-check the model name gemini-3.6-flash — typos return an error rather than a silent fallback

Authorizations

Authorization
string
header
required

Add Authorization: Bearer YOUR_API_KEY header

Body

application/json
model
string
required

Fixed: gemini-3.6-flash

messages
object[]
required

Standard OpenAI messages array

max_tokens
integer

Output quota; 2000+ recommended with thinking on (thinking bills as output)

temperature
number
stream
boolean

SSE streaming

reasoning_effort
enum<string>

Thinking tier

Available options:
low,
medium,
high
response_format
object

Structured output, json_schema supported

tools
object[]

Function-calling tool list

Response

Success

id
string
model
string
choices
object[]
usage
object

Usage stats