Skip to main content
POST
Text generation: Gemini 3.5 Flash-Lite (native Gemini format)
Use the playground on the right: put sk-your-api-key (your APIYI token — no Google Key needed) in x-goog-api-key and hit send — no thinking by default, so responses are fast.
No thinking output by default. For deep reasoning pass thinkingConfig: {"thinkingLevel": "high"} (only the high tier reliably triggers thinking in our tests). Search grounding, code execution, URL context, and Maps tools work only on this native endpoint. Capabilities and pricing: Gemini 3.5 Flash-Lite Overview.
  • For streaming, change the URL to :streamGenerateContent?alt=sse (the playground demonstrates non-streaming)
  • Code execution works, but executableCode / codeExecutionResult fields are currently not echoed — results appear in the text body
  • :countTokens and the explicit cache API are not yet enabled; Computer Use is not supported by this model officially

Parameter quick reference

Response notes

  • Thinking spend is in usageMetadata.thoughtsTokenCount (0 by default)
  • Grounding sources appear in candidates[0].groundingMetadata; URL context in urlContextMetadata
  • No implicit-cache hits observed for this model in our tests — don’t build cost models on them

Authorizations

x-goog-api-key
string
header
required

APIYI token, the sk- prefixed key

Body

application/json
contents
object[]
required

Conversation contents; parts can mix modalities

systemInstruction
object

System instruction

generationConfig
object

Generation config

tools
object[]

Tools: google_search / url_context / codeExecution / functionDeclarations, etc.

Response

Success

candidates
object[]
usageMetadata
object

Usage; thoughtsTokenCount is the thinking spend

modelVersion
string