Gemini 3.5 Flash Lite

Overview

This model is Google’s entry-level tier for 1M-token context windows. It accepts text, images, audio, and video as input, but generates text-only output. The architecture prioritizes throughput and low latency, making it suitable for batch processing and high-volume workloads. On Kyma, the model runs through an OpenAI-compatible endpoint with automatic request failover. It supports tool calling and prompt caching, which bills repeated prefixes at this model’s cached input rate. Every response includes exact usage costs in the usage.cost field and identifies the routed model via the X-Kyma-Model header. The model caps output at 8,192 tokens and does not support native structured outputs or explicit reasoning traces. Thinking tokens are included in the output pricing, so monitor token counts if your prompts trigger internal processing.

Specs

Field	Value
Model ID	`gemini-3.5-flash-lite`
Best for	High-volume classification, extraction and summarisation at 1M context
Context window	1.048576M
Max output tokens	8K
Input modalities	Text, Image, Audio, Video
Output modalities	Text
Tool calling	Yes
Structured outputs	Yes
Prompt caching	Yes
Speed	Fast
Cost band	Balanced
Release stage	Stable

Pricing

	Per 1M tokens
Input	$0.405
Output	$3.375

Use this when

High Volume Document Analysis — Process long reports or transcripts to extract key data points without hitting context limits.
Automated Content Categorization — Route large batches of user messages or logs into predefined labels using tool integration.
Audio And Video Summarization — Condense multimedia recordings and image sets into concise text summaries.
Fast Interactive Chat Routing — Handle lightweight conversational turns where response speed matters more than complex reasoning.

Not ideal for

Do not use this model for tasks requiring explicit reasoning chains, strict JSON schema enforcement, or outputs longer than 8,192 tokens.

Pick something else when

You need explicit reasoning or step-by-step logic: use deepseek-r1.
You require strict JSON schema enforcement: use gpt-5.6-terra or qwen3.7-flash.
You need outputs exceeding 8,192 tokens: use gemini-3.6-flash or qwen-3.7-max.

Example

from openai import OpenAI

client = OpenAI(base_url="https://kymaapi.com/v1", api_key="kyma-...")

response = client.chat.completions.create(
    model="gemini-3.5-flash-lite",
    messages=[{"role": "user", "content": "..."}],
    tools=[...],  # function calling supported
)

Agent query example

Ask the API which models fit, instead of hardcoding an id:

curl "https://kymaapi.com/v1/models?recommended_for=chat&tools=true&quality_tier=efficient"

FAQ

Does this model support structured JSON outputs natively? No, it does not support structured outputs. You will need to parse the generated text manually or route the request to a model with native schema enforcement. How does prompt caching affect my input costs? Kyma bills cached prompt prefixes at this model’s cached input rate, which is published beside the input rate rather than assumed. Can I send audio and video files directly to the API? Yes, the model accepts text, image, audio, and video inputs, though it only generates text responses.

Alias	Resolves to
`fast`	`gemini-3.5-flash-lite`

​Overview

​Specs

​Pricing

​Use this when

​Not ideal for

​Pick something else when

​Example

​Agent query example

​FAQ

​Related aliases