See how ChatGPT, Perplexity and Google AI Overviews describe you today
Free AI Visibility AuditGemma 3 4B is a small open-weight model from Google that takes both text and images as input and returns text. It is the compact member of the Gemma 3 family, built to run cheaply at scale rather than to win on hard reasoning.
GoogleVerified 24/09/2026Released 13/03/202501 / decision
Each figure sits next to the middle of the field, so it can be read as dear or cheap, wide or narrow, rather than floating on its own. Compared against the 179 models we track, not against an absolute standard.
Capability
Not measured
No published benchmark scores yet
Input / 1M tokens
$0.05
field median $0.46
Output / 1M tokens
$0.10
field median $1.82
Context
131K
field median 524K
02 / overview
Four questions, answered separately, because somebody arrives at one of them rather than at the top of the page.
The job it was built for, and the job it is not for.
Gemma 3 4B is a small multimodal language model: vision-language in, text out, with support for over 140 languages and improved maths, reasoning and chat behaviour over earlier Gemma releases. It is built for high-volume, well-defined work such as classification, extraction, tagging, summarising and routing. It is not the model to reach for when a task needs deep multi-step reasoning or long-form authored output.
Where it sits in its line, and when a sibling is the better pick.
Gemma 3 4B first appeared in March 2025 as part of the Gemma 3 launch, sitting at the small end of the family alongside larger Gemma 3 sizes. Pick it when the job is repetitive, the volume is high and the cost per call matters more than the last few points of quality. Pick a larger sibling when the task involves genuine reasoning or the output is going in front of a customer unedited.
The API, the apps it powers, and what its limits let you do.
Gemma 3 4B is available as open weights, so it can be self-hosted, and it is also served through hosted inference providers with a standard token-priced API. The context window is long enough to hold a full document set, a long support thread or a batch of records in a single call, and image input means screenshots, scanned pages and product photos can go in alongside the text. Output length is capped well below the input budget, which suits structured returns rather than long essays.
What changes because this exists, or why it does not.
Gemma 3 4B makes it reasonable to run a language model over every record rather than a sample. Work that used to be awkward, tagging an entire content archive, reading images across a whole product catalogue, screening thousands of inbound messages before a bigger model sees them, becomes a matter of throughput rather than budget. As a single-call model it is unremarkable, its value is in the volume it absorbs.
Follows Gemma 3 12B, and is the newest in its line. See the whole line.
03 / evidence
Split by category, so a strong number never hides a thin evidence base. Verified means we read it on the benchmark's own published results; a provider's claim about its own model is shown and labelled rather than dropped.
No published benchmark scores for this model yet.
We publish a score only where we can link the result to where it was published. Until a benchmark result for this model exists in a source we read, this section stays empty rather than being filled with a provider's marketing figure.
How we decide what counts as evidence04 / ledger
Every published row, grouped by category, each compared with the best published score on the same benchmark. 'Is 64% good' is a question nobody can answer; '26 points behind the leader' is one anybody can.
Nothing in the ledger yet.
Each row here carries a score, the benchmark it came from, what the leading model scored on the same test, and a link to the published result. Rows appear as results are published and read.
How we decide what counts as evidence05 / capability
Where this model is strong, and against how many peers. Ranks are against models with evidence in that category, not against everything we track: ranking against models nobody tested would rank who published, not who is better.
No category scores to shape yet.
A category score is the weighted mean of the benchmarks published for it. With no published rows there is nothing to average, and an empty chart drawn at zero would say something false.
How we decide what counts as evidence06 / cost
List API rates as last read from the provider, with the source on every row, plus every change we have recorded since we started tracking it.
| Charge | Price | Unit | Read on | Source |
|---|---|---|---|---|
| Input | $0.05 | per 1M tokens | 2026-09-24 | Check |
| Output | $0.10 | per 1M tokens | 2026-09-24 | Check |
Gemma 3 4B sits at the cheap end of the market, priced for bulk work where a call happens thousands of times a day rather than once. Output costs twice what input does, which is the usual shape, and self-hosting the open weights removes per-token cost entirely if you have the infrastructure.
| Workload | Input / month | Output / month | Cost |
|---|---|---|---|
| A small product team | 20M tokens | 5M tokens | $1.50 |
| A busy support assistant | 200M tokens | 40M tokens | $14.00 |
| A document pipeline | 1000M tokens | 100M tokens | $60.00 |
List API rates, no caching and no batch discount, which both providers offer and which change the answer a great deal. Treat these as the ceiling, not the bill.
07 / specs
As listed by the provider's own catalogue and re-read every few hours. Anything absent is absent there too.
| Context window | 131,072 tokens |
|---|---|
| Maximum output | 16,384 tokens |
| Modalities | text, image |
| Released | 13/03/2025 |
| Status | Current |
| Catalogue identifier | google/gemma-3-4b-it |
08 / lineage
What this model replaced, what replaced it, and what else its provider has in the field.
Also from Google
09 / line
Every model in this family in release order, so a page from eight months ago says in one glance that two newer ones exist.
Came before
← Gemma 3 12BCame after
Nothing newer in this line yet.
Ordered by release date and worked out from the naming, so a new member slots in as soon as its page exists. A retired model keeps its page and its place in the line.
10 / notes
What this model changes for a brand trying to be cited in AI answers, and every change we have logged since it launched.
What it changes for you
Gemma 3 4B does not answer buyer questions, so it will not cite your brand directly. What it changes is the pipeline behind the answers: models this cheap mean more of the web gets read, parsed and classified in bulk, including your documentation, pricing pages and support content. Structure that content so a small model can extract facts from it without inference, because increasingly a small model is what reads it first.
Buyers rarely meet Gemma 3 4B as a named product. They meet it inside software that embeds it, as the layer that classifies a ticket, extracts fields from an uploaded document or writes an image caption, and engineers meet it directly through the open weights or a hosted inference API. It is not the model behind a consumer chat assistant or an AI search surface.
Nothing published here yet. Changes appear within hours of a provider announcing them.
11 / questions
The things people ask about this model, answered from what is on this page rather than from anywhere else.
Models change what gets cited. We measure whether AI answers name your brand or your competitors across every assistant, and show you what to change.