Gemini 3.5 Flash Lite API

Access Google Gemini 3.5 Flash Lite—also searched as Gemini 3.5 Flash-Lite— through EvoLink's unified chat API. Test low-latency routing, structured output, tool use before integrating.

GoogleText GenerationAvailable

From $0.271 / 1M input tokens$0.300 official price-10%

API docs

Low-latency inferenceTool useStructured outputPrompt cachingChat + Gemini API

Production routeLive

Live

Context

1.05M context · 65.5K max output

Best For

Classification, extraction, routing, subagents

Input

Text + images

Output

Text · JSON (structured output) · tool calls

Models Pricing Readme API

Choose Gemini 3.5 Flash Lite

Google’s cheapest, fastest 3.5-class model for low-latency classification, routing, extraction, and lightweight agent subagents, with a full 1M-token context at $0.300 / $2.500 per 1M tokens.

Gemini 3.5 Flash Lite

Google fastest, most cost-effective 3.5-class model

Selected

Model ID

gemini-3.5-flash-lite

Best for

High-throughput classification and extraction, low-latency real-time actions, focused agent subagents, and cost-sensitive workloads where speed and price matter more than maximum reasoning.

Input

$0.271 / 1M-10%

18.4 cr / 1M$0.300official price

Cache read

$0.028 / 1M-7%

1.9 cr / 1M$0.030official price

Output

$2.250 / 1M-10%

153 cr / 1M$2.500official price

All rates are per 1M tokens, shown in USD and credits, and reflect your account's current pricing.

Gemini 3.5 Flash Lite pricing

Estimate what one Gemini 3.5 Flash Lite request costs before you integrate. The calculator uses your account's current rates, with official pricing as a reference.

Request calculator

Enter the token mix for one request and the number of Google Search queries.

Input tokensCache read tokensOutput tokensGoogle Search

Estimated request cost

Gemini 3.5 Flash Lite

USD$0.0010

Credits0.0647

Official estimate $0.0011 · save $0.0002 (10%)

Input tokens0.0184 cr

Cache read tokens0.0004 cr

Output tokens0.0459 cr

Google Search0 cr

Minimum charge: 0.01 credits per request.

Budget guide

Approximate requests using the current token mix and search queries.

Add credits

$10

About 10046 requests

For quick testing

$50

About 52550 requests

For regular development

$100

About 105100 requests

For production evaluation

Server-side tool rates

Gemini 3.x bills each nonempty search query, even if no sources are returned. One request may run multiple queries. Search fees are added separately after the token minimum charge.

Google Search$0.014/ query0.952 cr / query

Gemini 3.5 Flash Lite API for fast, low-cost, high-throughput workloads

Call Google’s cheapest, fastest 3.5-class model through EvoLink’s unified API. Gemini 3.5 Flash Lite is built for low-latency classification, routing, extraction, and lightweight agent subagents — with a 1,048,576-token context window, prompt caching, structured output, and tool use — at $0.300 / $2.500 per million input / output tokens.

Gemini 3.5 Flash Lite is served on EvoLink under the model ID gemini-3.5-flash-lite through Chat Completions · Gemini native API, with the same API key and balance you use for every other model. It offers a 1.05M context window and up to 65.5K output tokens, plus low-latency routing, structured output, tool use.

Gemini 3.5 Flash Lite specs and capabilities

Numbers come from the EvoLink route configuration; capabilities are what the API exposes today.

Context window: 1.05M tokens
Max output: 65.5K tokens
Input: Text + images
Output: Text · JSON (structured output) · tool calls
Tool use: Function calling with multi-step tool sequences
Prompt caching: Automatic cache reads at a lower rate
Server-side tools: Google Search, billed per search query
Protocols: Chat Completions · Gemini native API
Model ID: gemini-3.5-flash-lite

Why Gemini 3.5 Flash Lite can handle these workloads

Gemini 3.5 Flash Lite pairs a low price and high speed with a full 1M-token context and prompt caching. It is not a maximum-reasoning model; the value is doing simple work fast and cheap at scale.

A 1M-token context on the cheapest tier

The 1,048,576-token window lets even the cheapest 3.5-class model take long documents, transcripts, or logs in one pass. Retrieval and context compaction still reduce cost, since irrelevant input still consumes tokens.

Speed and throughput over deep reasoning

With about 350 output tokens/second and minimal thinking by default, Flash Lite optimizes for fast, high-volume responses. Reserve heavier reasoning configurations and models for the requests that actually need them.

Prompt caching lowers cost further

Stable system prompts, instructions, and tool schemas create cache hits that read at the dedicated lower cache rate. Keep prefix ordering consistent so repeated high-volume calls stay cheap.

Where Gemini 3.5 Flash Lite earns a place in a production model stack

Google positions Gemini 3.5 Flash Lite as its fastest, most cost-effective 3.5-class model. Its strongest fit is high-volume, latency-sensitive work where low price and speed matter more than maximum reasoning — and where you can escalate the few hard requests to a bigger model.

High-throughput classification and extraction

Gemini 3.5 Flash Lite is built for high-throughput classification, routing, and JSON extraction. At $0.300 / $2.500 per 1M tokens it lets you process large request volumes cheaply where a premium model would be wasted.

Low-latency, cost-sensitive scale

Google reports about 350 output tokens per second, making it a strong fit for real-time UI actions, autocomplete, moderation, and other latency-critical paths. Measure cost and speed per request, not headline benchmarks.

Lightweight and focused agent subagents

It suits subagents that execute focused tasks inside larger multi-agent systems. It defaults to minimal thinking for speed; for multi-step tool use, validate that agents do not terminate early before you scale.

When to escalate to a bigger model

Deep repository refactoring, long autonomous coding, and hard multi-step reasoning belong on Gemini 3.6 Flash or a Pro-tier model. Use Flash Lite for the high-volume majority and route only the difficult minority upward.

What early Gemini 3.5 Flash Lite reactions suggest—and what still needs proof

Gemini 3.5 Flash Lite launched on 2026-07-21 as a replacement for Gemini 2.5 Flash. Treat launch-day reactions as hypotheses to verify on your own tasks, tools, budgets, and acceptance criteria.

Price-to-performance is the headline

At $0.300 input / $2.500 output per 1M tokens and roughly 350 tokens/second, its appeal is throughput per dollar. On simple, high-volume work this can beat pricier models on total cost with acceptable quality.

It replaces Gemini 2.5 Flash and lighter 3-Flash workloads

Google frames it as a drop-in for Gemini 2.5 Flash and less complex Gemini 3 Flash tasks. If you run those today, re-benchmark your prompts before migrating so quality and formatting stay within tolerance.

Minimal thinking is the default—validate multi-step agents

Flash Lite defaults to minimal thinking for speed and cost. Google notes that minimal thinking can cause premature tool termination on multi-step tasks, so test agent completion before high-throughput rollout.

It is a closed, API-only Google model

Gemini 3.5 Flash Lite has no open weights and cannot be self-hosted. Access is through the Gemini API and platforms such as Google AI Studio, plus unified gateways like EvoLink—route by cost and latency, not local deployment.

Two ways to use Gemini 3.5 Flash Lite: EvoLink API or Agent

Use the EvoLink API for product backends and batch jobs, or call Gemini 3.5 Flash Lite from Codex, Claude, or Gemini for coding and analysis workflows. Both paths share the same EvoLink API key, balance, model ID, and request history.

Option 1

Integrate with the EvoLink API

Best for: product backends, batch jobs, automated pipelines

Send Gemini native generateContent requests to EvoLink and control the model ID, system instruction, output budget, tools, and structured output.

1Create an EvoLink API key in the console
2Call the generateContent endpoint on the EvoLink base URL with the model ID shown above in the request path
3Send one representative request and read usageMetadata for input, cached, and output tokens
4Set maxOutputTokens and retries per task; keep function-call parts and their results across turns

Create EvoLink API Key

Option 2

Call it with an Agent

Best for: coding, review, and analysis tasks in Codex, Claude, and Gemini

Give the Agent the task, the inputs to include, and the acceptance criteria. It assembles the request, calls Gemini 3.5 Flash Lite through EvoLink, and returns the answer with token usage.

1Set EVOLINK_API_KEY in your local environment; never put it in code or a prompt
2Describe the task, the inputs to include, and the expected output format
3Ask the Agent to call Gemini 3.5 Flash Lite through EvoLink and show the request before sending
4Let the Agent report the answer, token usage, and any error body

Gemini 3.5 Flash Lite API code example and error handling

This example shows the shortest runnable request: a Gemini native generateContent call with a system instruction, a user message, and an output budget. Open the API tab for the complete parameter and response reference.

View complete API docs

cURL

# Gemini native API
curl -X POST "https://api.evolink.ai/v1beta/models/gemini-3.5-flash-lite:generateContent" \
  -H "Authorization: Bearer $EVOLINK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "systemInstruction": { "parts": [{ "text": "You are a precise assistant. Answer in JSON when asked." }] },
    "contents": [
      { "role": "user", "parts": [{ "text": "Classify the sentiment of each review below and return a JSON array:\n<reviews>" }] }
    ],
    "generationConfig": {
      "maxOutputTokens": 1024,
      "temperature": 0.1
    }
  }'

# For streaming, call streamGenerateContent?alt=sse with the same body.
# The response includes candidates[0].content.parts and usageMetadata
# (prompt, cached, and output token counts).

Invalid request or unsupported parameter

Check the model ID, the contents array, and parameter ranges against the API reference; remove fields this route does not support.

Authentication or balance issue

Check the Authorization bearer token and confirm the available balance in the console.

Context length exceeded

Prompt tokens exceed the Gemini 3.5 Flash Lite context window. Trim or retrieve only the relevant evidence and reuse cached prefixes.

Rate limited (429)

Back off and retry with jitter; batch or queue requests instead of sending parallel bursts.

Content or tool call rejected

Review sensitive content, malformed tool-call arguments, and JSON schema mismatches before retrying.

What to verify before routing production traffic to Gemini 3.5 Flash Lite

A suitable workload can still fail because the integration uses the wrong identifier, sends unsupported parameters, or assumes deep reasoning that minimal-thinking defaults do not provide. Verify the request contract before evaluating quality.

Use the exact model ID gemini-3.5-flash-lite

Send model "gemini-3.5-flash-lite" (with a dot after 3.5) on the EvoLink API route. The dashed gemini-3-5-flash-lite is only the page URL, not the API model parameter.

Model ID

Use OpenAI Chat Completions or the Gemini native API

EvoLink exposes Gemini 3.5 Flash Lite through OpenAI-compatible /v1/chat/completions and the Gemini native generateContent endpoint. Keep the same EvoLink API key for either protocol.

Protocols

Mind the breaking parameter changes

Custom temperature, top-K, and top-P are ignored; custom frequency and presence penalties return an error; and a request whose last turn has the model role is rejected. Update older Gemini request builders accordingly.

Migration

Validate multi-step agents under minimal thinking

Because Flash Lite defaults to minimal thinking, confirm that multi-step tool-calling agents complete their steps rather than stopping early. Reserve harder reasoning for a larger model instead of over-tuning this one.

Agents

Prompt caching for repeated long context

Automatic cache reads use a dedicated lower rate when stable repository instructions, reference documents, and tool schemas can be reused.

1M-token context with a 65K default output limit

Keep connected code, specifications, documents, and agent state in one working context. Treat the limit as capacity rather than a target: retrieve relevant evidence and set task-appropriate output budgets.

Compare Gemini 3.5 Flash Lite cost per accepted task, not token price alone

Gemini 3.5 Flash Lite wins when its low price and speed clear your quality bar on high-volume work. Evaluate identical task sets, and escalate the requests it cannot handle rather than paying premium rates for all traffic.

First-pass successAccepted deliverable rateRetry countOutput tokensCache-hit rateTokens per secondCost per 1K requestsEscalation rate to a larger model

If Gemini 3.5 Flash Lite clears your quality bar on simple, high-volume tasks, its low rate and speed make it the cheapest production default. Route only the requests it fails to Gemini 3.6 Flash or a Pro-tier model.

Compare Gemini 3.5 Flash Lite with leading long-context models after workload testing

EvoLink

First verify whether Gemini 3.5 Flash Lite reduces retries and review effort on your tasks. Then compare price, context, caching, and workload fit to choose the production route.

Model	Gemini 3.5 Flash Lite	Gemini 3.6 Flash	Gemini 3.1 Pro
Input / output	$0.271 / $2.25	$0.675 / $3.375	$1.865 / $11.183
Context	1.05M	1.05M	1.05M
Caching	Automatic cache reads	Context cache	Context cache
Best for	High-throughput classification and extraction, low-latency real-time actions, focused agent subagents, and cost-sensitive workloads where speed and price matter more than maximum reasoning.	The workhorse to escalate to when a task needs stronger coding or reasoning than the Lite tier provides.	Step up to the Pro tier for the deepest reasoning tasks the Flash line is not built to handle.

Gemini 3.5 Flash Lite

Input / output$0.271 / $2.25

Context1.05M

CachingAutomatic cache reads

Best forHigh-throughput classification and extraction, low-latency real-time actions, focused agent subagents, and cost-sensitive workloads where speed and price matter more than maximum reasoning.

Gemini 3.6 Flash

Input / output$0.675 / $3.375

Context1.05M

CachingContext cache

Best forThe workhorse to escalate to when a task needs stronger coding or reasoning than the Lite tier provides.

Gemini 3.1 Pro

Input / output$1.865 / $11.183

Context1.05M

CachingContext cache

Best forStep up to the Pro tier for the deepest reasoning tasks the Flash line is not built to handle.

Gemini Model Family

Same API key and balance — switch tiers without changing your integration.

Compare all Gemini models

Gemini 4 Argon

Google’s Gemini 4 model, with a 1M-token output limit and a lower rate for cached input.

View model

Gemini 3.8 Flash

The newest Flash generation for stronger coding, reasoning, and agent workloads.

View model

Gemini 3.7 Flash

A recent Flash generation for strong reasoning at a fast, efficient tier.

View model

Other text models on EvoLink besides Gemini 3.5 Flash Lite

GPT-5.6

OpenAI’s tiered frontier family (Sol/Terra/Luna) for capability, latency, and cost-routing flexibility.

View model

Claude Opus 4.8

Anthropic’s premium route for long-running agents, complex reviews, and judgment-heavy work.

View model

DeepSeek V4

DeepSeek’s cost-efficient 1M-context route for high-volume coding, reasoning, and agent workloads.

View model

Kimi K3

Moonshot’s long-context reasoning route for repository-scale coding and multi-document work.

View model

Gemini 3.5 Flash Lite guides and related reading

Compare the Gemini API family

See how 3.5 Flash-Lite, 3.6 Flash, 3.5 Flash and 3.1 Pro differ on price, speed, context, and workload fit before you route.

Read guide

Gemini 3.6 Flash API

The workhorse to escalate to when a task needs stronger coding or reasoning than the Lite tier provides.

Read guide

Gemini 3.5 Flash API

The mid Flash tier between Flash-Lite and the Pro line, for a balance of cost and capability.

Read guide

Gemini 3.1 Pro API

Step up to the Pro tier for the deepest reasoning tasks the Flash line is not built to handle.

Read guide

Gemini 3.5 Flash Lite API FAQ

Is the Gemini 3.5 Flash Lite API available through EvoLink?

Yes. Gemini 3.5 Flash Lite is available as a production model with live backend pricing and official fallback rates.

What model ID should I use for the Gemini 3.5 Flash Lite API?

Send model "gemini-3.5-flash-lite" (with a dot after 3.5). The dashed gemini-3-5-flash-lite is only the page URL, not the API model parameter.

Which protocols and SDKs work with Gemini 3.5 Flash Lite?

Use the OpenAI-compatible Chat Completions endpoint or the Gemini native generateContent endpoint with the same EvoLink API key. There is no Anthropic Messages route for this model.

Does the Gemini 3.5 Flash Lite API support 1M context or only 256K?

The EvoLink route records 1,048,576 tokens. Smaller limits usually come from a particular Gemini product surface or client configuration, not the API model itself.

How fast and how cheap is Gemini 3.5 Flash Lite?

It is Google’s fastest, most cost-effective 3.5-class model — $0.300 input / $2.500 output per 1M tokens at roughly 350 output tokens per second. That throughput per dollar is its main advantage.

What is Gemini 3.5 Flash Lite best for?

High-throughput classification, routing, JSON extraction, low-latency UI actions, and focused agent subagents. For deep coding or hard reasoning, escalate to Gemini 3.6 Flash or a Pro-tier model.

How is Gemini 3.5 Flash Lite different from 3.5 Flash and 3.6 Flash?

Flash Lite is the cheapest, fastest tier and defaults to minimal thinking. 3.5 Flash and 3.6 Flash are more capable, higher-priced workhorses for coding and agents.

Does Gemini 3.5 Flash Lite replace Gemini 2.5 Flash?

Google positions it as a replacement for Gemini 2.5 Flash and less complex Gemini 3 Flash workloads. Re-benchmark your prompts before migrating.

What breaking changes should I handle when migrating?

Custom temperature, top-K, and top-P are ignored; custom frequency and presence penalties return an error; and a request whose last turn has the model role is rejected.

Is Gemini 3.5 Flash Lite a good default for real-time or high-volume requests?

Yes — that is what it is built for. It defaults to minimal thinking for speed, so validate that multi-step agents complete their steps before a high-throughput rollout.

How should I compare Gemini 3.5 Flash Lite with GPT, Claude, GLM, or DeepSeek?

Compare it against other cheap, fast tiers on cost per 1K requests and latency, not against premium reasoning models. Escalate the requests it cannot handle to a larger model.

What should a production Gemini 3.5 Flash Lite evaluation measure?

Track cost per 1K requests, tokens per second, first-pass success, accepted deliverables, cache hits, and how often requests must escalate to a larger model.