Gemma 3 4B
Google · released Feb 20, 2025 · google/gemma-3-4b-it
- Type
- Open weightsGemma Terms of Use
- Params
- 4.3B
- Context
- 131K
about 98K words of context · download allowed, licence restricts use
Our take
Written Sep 30, 2026Gemma 3 4B is a small downloadable model that takes text and pictures and returns written answers, and its licence puts conditions on commercial use and redistribution. Its measured quality sits near the bottom of the field on every board we hold, so treat it as a cheap workhorse for low-stakes jobs rather than a model to build a product's quality on.
Use it for simple text and image tasks where a low per-token rate matters more than answer quality, or when you want to run a small model on your own hardware. Read the Gemma Terms of Use before you build anything commercial on it. Skip it if your task needs measured coding, reasoning or instruction-following quality, or if you need a placing above the bottom of the field.
The case for it
- Text and images go into the same request, so a screenshot does not have to be described in words first.
- You can download it and run it yourself, so a single modern graphics card is enough to try it without a host.
- All three listed offers sit at the same low rate, so a high-volume, low-stakes workload is cheap to run through a host.
The case against it
- Near the bottom of the field on every board we hold: 165th of 168 on Arena Coding and 146th of 168 on Arena Text, both as of 25 Sep 2026. These are human-preference rankings, not correctness scores.
- The Gemma Terms of Use put conditions on commercial use and redistribution, so the terms need reading before you build on it.
- Released 2025-02-20, so it is competing against newer small models.
How good is it?
A small open text model for basic chat, though it trails most models on everyday questions, writing and coding.
- getting answers to everyday questionsArena Text (overall) · 146th of 168
- drafts, rewrites and editingArena Creative Writing · 140th of 168
- writing and completing codeArena Coding · 165th of 168
EverydayGeneral questions and everyday reasoning
Arena Text (overall)146th of 168 · 1303
CodingWriting and fixing code on its own
Arena Coding165th of 168 · 1274
Arena Coding is the only board that has scored it for this.
AgenticPlanning, calling tools, staying on task
Not yet scored on Arena Agent.
WritingDrafting and rewriting prose
Arena Creative Writing140th of 168 · 1274
Arena Creative Writing is the only board that has scored it for this.
These tests check whether a model follows instructions — a precondition for all the work above, but not a measure of how well that work is done.
Every published score for this model6 scoresEvery figure we hold, from 6 boards, with who ran it and a link to the source — including the boards no rating above is built on.
Can you run it yourself?
Comfortable fit
GeForce RTX 4090 · 24 GB
Room to spare. 18.7 GB spare means a 10% error in the size would not change the answer.
GeForce RTX 5090 · 32 GB
Room to spare. 26.7 GB spare means a 10% error in the size would not change the answer.
Apple M2 (8-core GPU, 8GB unified) · 8 GB
Room to spare. 1.9 GB spare means a 10% error in the size would not change the answer.
Memory use by level
Against a 24 GB card.
What is quantisation? →This model on every device we track71 devicesThe Q4 build most people download, on each device: what the weights come to, how much context the memory leaves, and whether it runs. Smallest device that runs it first.
Check against your own machine → · Where to rent it hosted →
Or rent it from someone else
Prices checked 2 hours ago — each listing carries its own date.
- per 1M tokens
- $0.050 in / $0.10 out
- Context served
- 131K
- Throughput
- Not measured
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| OpenRouterOpenRouter's own listing | $0.050 / $0.10checked 2 hours ago | 131K | not measured | Unknown | Unknown | Unknown |
| DeepInfrabf16Direct and through OpenRouter | $0.050 / $0.10checked 2 hours ago | 131K16K max reply through OpenRouter | 17 tok/sthrough OpenRouter | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterConfirmed |
DeepInfraAcross the 2 listings we hold: 1 says it does not train on prompts (only through OpenRouter), 0 say they do and 1 does not say. 1 appears in the zero-retention registry we check (only through OpenRouter); the rest are unknown to us.
What each host's API supports
From the parameter list each endpoint publishes. Streaming is omitted: nothing we hold reports it, for any model.
| Provider | Tool calling | JSON output | Strict schema |
|---|---|---|---|
| OpenRouterOpenRouter's own listing | ✗ | ✓ | ✓ |
| DeepInfrabf16Direct and through OpenRouter | ✗ | ✓ | ✓ |
Tool calling: 0 of 2 listings say yes, 2 say no. JSON output: 2 of 2 listings say yes. Strict schema: 2 of 2 listings say yes.
Models people weigh against Gemma 3 4B
When we formed this view
Recent changes
What moved
first indexed by our pipelineEach date is the day we first saw the change, or the day the maker announced it.
What we do not know about this model yet
- We hold no measured file for it, so all 3 sizes on this page are calculated from the parameter count.
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 1 of 2 listings does not say whether it trains on prompts, and 1 answers only through OpenRouter, not for its own listing.
- We hold no cached-input rate for any of its listings.
- We hold no batch or off-peak rate for any of its listings.
- We hold a decode speed for it, but no prompt-processing (prefill) figure, so how long the input side of a job takes is unknown to us.
Licence and identifiers
What the licence allowsGemma Terms of Use, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.
Licence
Gemma Terms of Use
Commercial use allowed, but Google's prohibited-use policy applies and can be updated over time — terms are less static than Apache/MIT.
Identifiers
- Hugging Face
- google/gemma-3-4b-it
- Takes in, gives back
- Text and images in, text out
- Catalogue slug
- google-gemma-3-4b