← Model indexMODEL PROFILE / gemini-3-8-flash

Google / current

Gemini 3.8 Flash

Google's high-speed multimodal tier tuned for low latency and high output throughput.

Context1M
ModalityText · Image · Video · Audio
AccessAPI · Cloud
ReasoningControllable

Where it fits.

Keeps very high token output speed while retaining complex agentic, coding and multimodal capabilities. The current Flash generation continues Gemini 3.x migration guidance: move to thinking_level, drop deprecated sampling fields, and rely on explicit model IDs over latest aliases.

02

Core capabilities

  • Very high output throughput relative to its class
  • Text, image, video and audio input
  • Thinking tiers, code execution and search grounding
  • Function calling and structured output
03

Best-fit workflows

  • Live chat and support
  • High-volume multimodal parsing
  • Latency-sensitive agents
04 / OPERATIONS

From evaluation
to production.

Deployment notes

Default choice when tail latency dominates. Use the stable ID rather than the latest alias, keep thinking at the lowest level that preserves quality, and watch.elevate to Argon for deep reasoning cases.

Watch before choosing

Speed-focused tiers trade some depth; validate exact joint behaviour when migrating older scripts. Deprecated sampling parameters still silently change behaviour on some surfaces; remove them explicitly.

05

Family & versions

Specifications and positioning checked against provider documentation.

Open official source↗