See how ChatGPT, Perplexity and Google AI Overviews describe you today
Free AI Visibility AuditGemma 4 31B is Google's open-weight multimodal instruct model, a roughly 30.7 billion parameter dense model that reads text, images and video and writes text back. It sits in the Gemma family as the option you run yourself or call cheaply through a host, rather than a hosted frontier model.
GoogleVerified 24/09/2026Released 02/04/202601 / 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 231 models we track, not against an absolute standard.
Capability
Not measured
No published benchmark scores yet
Input / 1M tokens
$0.09
field median $0.43
Output / 1M tokens
$0.34
field median $1.80
Context
262K
field median 500K
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 4 31B is a dense instruction-tuned multimodal model built for general text generation with image and video input alongside it. Google DeepMind ships it with a configurable thinking mode and native function calling, so it is aimed at agent and tool-use work as much as plain chat. It is not a frontier reasoning model and it is not a hosted assistant product, it is a component you deploy into your own stack.
Where it sits in its line, and when a sibling is the better pick.
Gemma 4 31B first appeared in April 2026, positioned in the middle of the Gemma 4 line, above the small models meant for on-device work and below Google's closed Gemini tier. Pick it when you want open weights, multimodal input and function calling without paying hosted frontier rates. It is the wrong pick when a task genuinely needs the strongest available reasoning, or when you need very long generated outputs in a single pass.
The API, the apps it powers, and what its limits let you do.
Gemma 4 31B is reached through the usual API endpoints of whichever provider hosts it, or self-hosted from the open weights. The quarter-million token context comfortably holds a full documentation set, a long transcript or a batch of product pages in one call, and image and video input means you can feed screenshots, interface recordings or product photography without a separate vision step. Output is capped well below the input window, so it suits analysis, extraction and structured responses rather than long-form drafting in one shot.
What changes because this exists, or why it does not.
An open-weight model that takes video and images, holds a very large context and calls functions natively removes a lot of the plumbing teams used to build around smaller open models. Work that previously meant a vision model, a text model and a separate orchestration layer can now sit in one call, at a cost low enough to run across a whole content library rather than a sample of it.
Follows Gemma 3 4B. Superseded by Gemma 4 26B A4B. 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 | $0.09 | per 1M tokens | 2026-09-24 | Check |
| Output | $0.34 | per 1M tokens | 2026-09-24 | Check |
Gemma 4 31B is priced at the low end of the market, cheap enough that running it across large document sets or every page on a site is a routine decision rather than a budgeted one. Output costs a few times more than input, as usual, but both sit far below hosted frontier models, 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 | $3.50 |
| A busy support assistant | 200M tokens | 40M tokens | $31.60 |
| A document pipeline | 1000M tokens | 100M tokens | $124.00 |
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 | 262,144 tokens |
|---|---|
| Maximum output | 16,384 tokens |
| Modalities | image, text, video |
| Released | 02/04/2026 |
| Status | Current |
| Catalogue identifier | google/gemma-4-31b-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 4BCame after
Gemma 4 26B A4B →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
Cheap open-weight multimodal models mean more companies will run their own retrieval and summarisation over your site rather than relying on a handful of big assistants. That pushes the burden back onto your source material being clean, complete and machine-readable, because there is no single AI engine whose quirks you can optimise for. Assume your pages, docs and screenshots are being read directly, at scale, by systems you will never see.
Buyers rarely meet Gemma 4 31B by name. They meet it inside products built on top of it, support assistants, content tools and internal search built by teams who wanted open weights and predictable costs, and through the model catalogues of the inference providers that host it.
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.