Gemini 3.5 Flash-Lite API
NewLLMGemini 3.5 Flash-Lite is the newest low-cost Gemini route: 1M context, vision, and tool use at a Lite rate.
- Input · USD / 1M tokens$0.255 USDOfficial price
$0.3 USD - Output · USD / 1M tokens$2.125 USDOfficial price
$2.5 USD - Savings15%
- Context1M
- View tool pricing and billing rules
What are we making today?
Bring an idea, a question, or a first draft.
01
What this model is for
Gemini 3.5 Flash-Lite is the newest low-cost Gemini route: 1M context, vision, and tool use at a Lite rate.
Gemini 3.5 Flash-Lite is the newest Lite-tier model Google offers in the Gemini line, listed on OmniaKey as gemini-3.5-flash-lite. It keeps the family's 1M-token context window, image input and tool use, adds the catalog's fast label, and names Cursor, Cline and Aider for client setup. Three rates apply: input, output and cached input. All three sit in a lower band than Gemini 3.8 Flash, which is the mainline model one step above it.
The point of a Lite id is arithmetic on volume. When a job is a short prompt with a short answer and it runs a hundred thousand times a day, the difference between the Flash band and the Lite band decides whether the feature ships. That trade holds for classification, tagging, field extraction, short summaries and first-pass screening. It does not hold for reasoning-heavy work: a task that needs a long chain of thought belongs on a Pro model, and one that needs many tool turns usually belongs on the mainline Flash entry.
A useful pattern is two passes. Let 3.5 Flash-Lite answer everything it can with a strict output format, mark low-confidence results with an explicit rule, and hand only those to Gemini 3.8 Flash or Gemini 4 Argon. The Lite pass keeps the common case inexpensive, and the escalation keeps quality where it matters. Log which pass produced each answer so a later review can tell whether the cheap pass was enough.
Best-fit coding workloads
Ticket tagging at volume
Assign two or three labels and a priority to each incoming support ticket, with a fixed list of allowed labels and a JSON answer. The Lite rate band makes a large queue affordable, and the 1M window means a long thread with attachments still fits. Practical tip: reject any label outside the list and count those rejections.
Invoice field extraction
Read a scanned invoice image and return supplier, date, currency, net and tax as one small object. Vision input covers the scan, and the short answer keeps the output charge low. Practical tip: require null instead of a guess when a field is unreadable, then send those documents to a model in a higher band.
Cheap first pass, then escalate
Screen a long document set for the few pages that mention a clause, then pass only those pages to a Pro model for the actual reading. Two passes cost less than one Pro pass over everything and often answer faster in practice. Practical tip: keep the screening prompt and the reading prompt in one file so a change to either is visible.
02
Hands-on notes
Checked on 2026-10-27 through the OmniaKey API with one standard coding prompt (median of 3 runs) and three capability probes.
- Output tokens for our standard coding prompt, reasoning included
- 318
- Cost of that output at current prices
- $0.0007
- JSON mode
- Passed
- Tool calling
- Passed
- Image input
- Passed
03
Technical specification
Catalog facts and integration details verified for this route.
- Model ID
- gemini-3.5-flash-lite
- Context
- 1M
- Inputs and tools
- Vision · Tool use · Fast responses
- API endpoint
- /v1/chat/completions
- Cache hit / 1M
- $0.0255 USD
- Coding clients
- Cursor · Cline · Aider
Limits to plan around
- It sits above Gemini 2.5 Flash-Lite in the catalog, so it is not the cheapest Gemini listing. A very cost-sensitive job should compare the two before assuming the newer Lite tier is the smaller bill.
- The catalog lists vision, tool use and the fast label, and lists no thinking-level or reasoning parameter for this id. Do not assume the mainline Flash thinking controls apply; set what is available and check the response.
- Lite models handle short, well-specified tasks. A prompt that needs multi-step planning, careful counting or long tool chains is the wrong shape for this tier, whatever the window size suggests.
- The catalog shows one rate band and no batch mode. Bulk work still needs your own queue, retry policy and rate control, and a failed item has to be caught by your validator rather than by the platform.
04
Compare nearby models
Compare workload fit and current listed token rates before choosing a route.
| Model | Best for | Context | Input | Output |
|---|---|---|---|---|
| Gemini 3.5 Flash-LiteThis page gemini-3.5-flash-lite | High-volume classification, extraction and screening where short answers and a small bill decide the design. | 1M | $0.255 USD Official price $0.3 USD | $2.125 USD Official price $2.5 USD |
| Gemini 3.8 Flash gemini-3.8-flash | The mainline Flash entry for agent loops that need more careful answers at a higher rate band. | 1M | $0.45 USD Official price $0.75 USD | $2.25 USD Official price $3.75 USD |
| Gemini 2.5 Flash-Lite gemini-2.5-flash-lite | The lowest-cost Gemini listing, for jobs that the 2.5-generation Lite model already handles well. | 1M | $0.085 USD Official price $0.1 USD | $0.34 USD Official price $0.4 USD |
05
SDK quickstarts
Use one base URL and model ID from cURL, Python, or TypeScript.
curl https://api.omniakey.com/v1/chat/completions \
-H "Authorization: Bearer $OMNIAKEY_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gemini-3.5-flash-lite",
"messages": [
{"role": "user", "content": "Review this pull request"}
],
"stream": true
}'06
Estimate token cost
Enter expected input and output tokens to compare current listed rates.
- OmniaKey estimate
- $0.068
- Official-rate estimate
- $0.08
- Estimated savings
- $0.012
Estimate covers input and output tokens only. Cache usage, retries, tools, and client-side token accounting can change the final charge.
07
Related coding guides
Deeper selection, setup, and troubleshooting guidance for this model family.
Gemini 3.8 Flash API Pricing and Agent Cost
Google's introductory rate is $0.75 input and $3.75 output per million tokens, but retries and thinking decide the cost of an accepted result.
Read guideGemini 4 Argon vs GPT-6 Astra: Same Tasks, Speed and API Cost
Three real jobs, the same prompts: who got it right, how fast, what it cost.
Read guide08
Sources and verification
Model facts and links were checked on 2026-10-27. Provider documentation remains the authority for upstream behavior.
Frequently asked questions
How much does Gemini 3.5 Flash-Lite cost on OmniaKey?
OmniaKey charges $0.255 USD per 1M input tokens and $2.125 USD per 1M output tokens, in USD. Official prices are $0.3 USD and $2.5 USD; you save 15%.
What is the context window for Gemini 3.5 Flash-Lite?
The listed context window is 1M. Actual usable limits can also depend on the client and upstream request format.
How is Gemini 3.5 Flash-Lite billed?
Three rates apply: one for input tokens, one for output tokens and a lower one for cached input. They sit in a band below Gemini 3.8 Flash, and the catalog shows a single band rather than tiers by prompt length. Current figures are in the price table on the model page.
What is the difference between Gemini 3.5 Flash-Lite and Gemini 3.8 Flash?
Rate band, mainly. Both list a 1M-token context window, image input, tool use, the fast label and the same three coding clients, so the capability list does not separate them. The Flash-Lite entry is the cheaper of the two, and the mainline Flash entry is the one to use when answers need more care.
Is this the cheapest Gemini model on OmniaKey?
No. Gemini 2.5 Flash-Lite lists lower rates on all three lines, and Gemini 2.5 Flash sits at the same level as this model. Compare the two Lite ids on your own prompts: the newer generation is not automatically the smaller bill, and answer quality decides how many results you have to redo.
How do I call Gemini 3.5 Flash-Lite?
POST to https://api.omniakey.com/v1/chat/completions with gemini-3.5-flash-lite as the model. Text, image parts and tool definitions all go in the messages array of an OpenAI-compatible body. The id is case-sensitive and includes both dots, so copy it from the model page rather than retyping it.
Can it read images and call tools?
The catalog lists both vision and tool use, and names Cursor, Cline and Aider. In practice, pair images with a narrow question, such as a single field on a form, because a small model does better with one clear request than with an open-ended reading task. Test your own scans before promising accuracy.