by Google DeepMind

Gemma 3 4B.

multimodal open weights edge 4B params 128K ctx

🧬 Distilled from Gemma 3 27B — smaller, cheaper to run, similar reasoning style.

Cheapest input

$0.04/M

on OpenRouter

Cheapest output

$0.08/M

on OpenRouter

Fastest

20 tok/s

on OpenRouter

Smallest GPU

1× Nvidia Titan V

4B Gemma 3 — laptop multimodal.

Smallest GPU to run it See all quantisations →

1× Nvidia Titan V.

Most-aggressive quantisation we have a working recommendation for. Lower precision = less VRAM = cheaper hardware, at a small accuracy cost.

Where to use it

Cheapest hosted endpoints.

Provider	Access	$/M in	$/M out
OpenRouter	api aggregator	$0.04	$0.08	Launch ↗

Performance

Speed across providers.

Tokens/sec and time-to-first-token measured against the same prompt template on each provider's API.

Provider	Tokens/sec	TTFT	Total
OpenRouter	20.4	920 ms	5921 ms

Sources

Official references.

Hugging Face ↗ GitHub ↗

Family

Variants in the Gemma family.

Gemma 3 27B

27.0B open Gemma Terms of Use

Google's open-weight multimodal LLM — efficient and license-permissive.

Gemma 3 12B

12.0B open Gemma Terms of Use

12B Gemma 3 — multimodal, single-GPU target.

Gemma 3 1B

1.0B open Gemma Terms of Use

1B Gemma 3 — edge / mobile.

Best for

Workloads.

Run LLMs (inference / serving)

FAQ

Frequently asked.

How do I run Gemma 3 4B?

Gemma 3 4B is open-weight, so you can self-host on rented GPUs. See the Run It Yourself tab for GPU configurations + cost estimates, or use one of the hosted inference providers listed on this page.

Where can I access Gemma 3 4B?

Gemma 3 4B is available via OpenRouter, Ollama. Each access option lists its own pricing (per million tokens or hourly hosting).

How much does it cost to run Gemma 3 4B?

Self-hosting cost depends on the GPU you rent and the throughput you need. See the Run It Yourself tab for GPU configurations and hourly cost estimates.

Is Gemma 3 4B open-source or proprietary?

Gemma 3 4B is open-weight under the Gemma Terms of Use license. You can download and self-host it.

Run it yourself

Cheapest hardware per quantisation.

Each row is one quantisation tier (the same weights compressed differently). Lower precision → lower VRAM → cheaper hardware, at the cost of small accuracy loss. $/hr refreshed hourly from each provider's API.

Quantisation	Cheapest GPU config	Total VRAM	Live $/hr	tokens/sec
FP16 FP16 — half precision (default)	1× Nvidia RTX 3080	10 GB	—	—	Compare →
FP8 FP8 — 8-bit float (Hopper / Blackwell)	1× Nvidia Nvidia P102-100	5 GB	—	—	Compare →
INT4 INT4 — 4-bit integer (~4× VRAM saving)	1× Nvidia Nvidia Titan V	3 GB	—	—	Compare →

Just need an API?

Skip the GPU rental and call a hosted endpoint instead. See access providers on the Overview tab →

API pricing

What it costs per month across providers.

Estimate your monthly bill for Gemma 3 4B across every host that publishes per-token pricing. Slide your token volumes; the chart + table re-rank cheapest-first.

Cheapest

$0.56

OpenRouter

$/M input

$0.04

per million tokens

$/M output

$0.08

per million tokens

Providers

with priced rows

Monthly bill

Cheapest provider on the left.

Total monthly cost — input + output tokens combined.

Per provider

Bill breakdown.

Provider	$/M in	$/M out	Input cost	Output cost	Monthly total
OpenRouter	$0.04	$0.08	$0.4	$0.16	$0.56	Sign up ↗

Full calculator

Want to compare token volumes across other models too? Open the standalone API pricing tool →

Gemma 3 4B.

1× Nvidia Titan V.

Cheapest hosted endpoints.

Speed across providers.

Official references.

Variants in the Gemma family.

Workloads.

Frequently asked.

Cheapest hardware per quantisation.

What it costs per month across providers.

Cheapest provider on the left.

Bill breakdown.

How much it can remember.

What it can do.