What is Gemma 4 31B good for?

Self-hosted text and image work at the heavy end of the Gemma 4 dense models. Apache 2.0 weights, tuned to fit a single consumer GPU.

When is Gemma 4 31B not a good fit?

Audio or video input, and hosted tools like web search or code execution. It is the heaviest Gemma 4 model to serve, and no maximum output is published.

What is the cheapest way to run Gemma 4 31B?

Hosted, unless you push serious volume. Geodd charges $0.13 in / $0.37 out per 1M tokens. The cheapest rental that fits is 2x RTX 4070 at $130 a month, which costs the same as about 762M tokens a month on that API.

Can I self-host Gemma 4 31B?

Yes. The weights are Apache 2.0 licensed. At 4-bit it needs about 22 GB of GPU memory, which starts at roughly $130 a month on the cheapest rental that fits.

Open weights · Apache 2.0

Gemma 4 31B

The largest dense model in Google's Gemma 4 family at 30.7B parameters, tuned for consumer GPUs. Apache 2.0 license.

Jump to: Specs Pricing Self-hosting Similar models FAQ More from Google Cloud

Visit website Compare vs other LLMs →

Cheapest via Geodd

Input: $0.13 / 1M tokens
Output: $0.37 / 1M tokens

5 providers

Key Specifications

Context window: 256K tokens
Knowledge cutoff: Jan 2025
Released: Apr 2026
Parameters: 30.7B
Inputs: Text, image

Capabilities Show details: Reasoning Thinks before it answers, always on or as a switchable mode. Function calling Connect to external tools, APIs, and systems. Structured output Return responses in structured formats like JSON.

Hosted API pricing

Input tokens M Output tokens M

Provider	Input / 1M tokens	Output / 1M tokens	Cost at 10M in + 2M out
Geodd Cheapest	$0.13	$0.37	$2.04	View
Novita	$0.14	$0.40	$2.20	View
Together AI	$0.39	$0.97	$5.84	View
EcoHash	$0.50	$0.50	$6.00	View
Google Cloud Creator	Open weights, no hosted price			View

Heads up: Base-tier, on-demand rates per 1M tokens; cached, batch and long-context tiers excluded. A provider may serve a shorter context or a quantized build than the creator's release. Verify before provisioning. More on how we price.

Estimated cost to self-host

Gemma 4 31B needs about 22 GB of GPU memory at 4-bit with 32K context. The cheapest rental that fits is 2× RTX 4070 at about $130 a month.

That costs the same as roughly 762M tokens a month on Geodd's API. Self-hosting is more expensive below that volume.

Precision	Cheapest, 32K context	Cheapest, full 256K context
4-bitINT4 / FP4	2× RTX 4070 22 GB VRAM · $130/mo	4× RTX 4070 40 GB VRAM · $259/mo
8-bitFP8 / INT8	4× RTX 3060 38 GB VRAM · $144/mo	4× RTX 5060 Ti 54 GB VRAM · $288/mo
16-bitFP16 / BF16	8× RTX 3060 69 GB VRAM · $288/mo	8× RTX 5060 Ti 105 GB VRAM · $576/mo

Estimates based on median on-demand rates for Nvidia GPUs. Memory is weights plus KV cache for one request, using FP8 KV cache where supported and FP16/BF16 otherwise. Break-even assumes a 5:1 input-to-output ratio. No guarantee of runtime support, usable performance, or that a matching quantized build exists. How we estimate costs.

Similarly priced models

The models nearest Gemma 4 31B by blended rate, each at its own cheapest provider.

Model	Blended / 1M	Input / 1M	Output / 1M	Context	Cutoff	vs Gemma 4 31B
Ministral 3 8B Mistral	$0.15	$0.15	$0.15	256K		−12%
Qwen3.5-Flash Alibaba Cloud	$0.15	$0.10	$0.40	1M		−12%
Llama 3.2 11B Vision Meta	$0.155	$0.05	$0.68	128K	Dec 2023	−9%
Nemotron 3 Super 120B A12B Nvidia	$0.1583	$0.09	$0.50	1M	Jun 2025	−7%
DeepSeek V4 Flash DeepSeek	$0.1633	$0.14	$0.28	1M		−4%
Gemma 4 31B This model Google Cloud	$0.17	$0.13	$0.37	256K	Jan 2025
Qwen3-235B-A22B Instruct (2507) Alibaba Cloud	$0.1717	$0.09	$0.58	262K		+1%
Llama 3.3 70B Meta	$0.175	$0.13	$0.40	128K	Dec 2023	+3%
Hy3 Tencent	$0.1967	$0.13	$0.53	262K		+16%
Ministral 3 14B Mistral	$0.20	$0.20	$0.20	256K		+18%
Qwen3.8-Flash Alibaba Cloud	$0.2033	$0.15	$0.47	1M		+20%

Prices are USD per 1M tokens at each model's cheapest listed provider. Blended is the cost of 10M input plus 2M output tokens, spread over the 12M.

Vast.ai

Our sponsor

Per-second billing. Pay only for what you use.

50+ GPU models from RTX 3060 to B200, from $0.04/hr

No minimums, no long-term contracts.

SOC2 certified · PyTorch, CUDA & TensorFlow templates ready

Learn more

Frequently Asked Questions

What is Gemma 4 31B good for?: Self-hosted text and image work at the heavy end of the Gemma 4 dense models. Apache 2.0 weights, tuned to fit a single consumer GPU.
When is Gemma 4 31B not a good fit?: Audio or video input, and hosted tools like web search or code execution. It is the heaviest Gemma 4 model to serve, and no maximum output is published.
What is the cheapest way to run Gemma 4 31B?: Hosted, unless you push serious volume. Geodd charges $0.13 in / $0.37 out per 1M tokens. The cheapest rental that fits is 2x RTX 4070 at $130 a month, which costs the same as about 762M tokens a month on that API.
Can I self-host Gemma 4 31B?: Yes. The weights are Apache 2.0 licensed. At 4-bit it needs about 22 GB of GPU memory, which starts at roughly $130 a month on the cheapest rental that fits.

More from Google Cloud

Model	Context	Cutoff	Input / 1M	Output / 1M
Gemini 2.5 Flash-Lite	1M	Jan 2025	$0.10	$0.40
Gemma 4 26B A4B Open weights	256K	Jan 2025	$0.10	$0.30
Gemma 3 27B Instruct Open weights	128K	Aug 2024	$0.10	$0.30
Gemini 3.1 Flash-Lite	1M	Jan 2025	$0.25	$1.50
Gemma 3 12B Instruct Open weights	128K	Aug 2024	$0.05	$0.10
Gemini 3.5 Flash-Lite	1M	Mar 2026	$0.30	$2.50
Gemini 2.5 Flash	1M	Jan 2025	$0.30	$2.50
Gemini 3 Flash Preview	1M	Jan 2025	$0.50	$3.00