Google: Gemma 3 12B.
Gemma 3 introduces multimodality, supporting vision-language input and text outputs. It handles context windows up to 128k tokens, understands over 140 languages, and offers improved math, reasoning, and chat capabilitie
1× Nvidia GeForce RTX 3060Ti LHR.
Most-aggressive quantisation we have a working recommendation for. Lower precision = less VRAM = cheaper hardware, at a small accuracy cost.
Cheapest hosted endpoints.
Variants in the Gemma family.
Gemma 3 27B
Google's open-weight multimodal LLM — efficient and license-permissive.
Gemma 2 27B
Google's pre-Gemma-3 open-weight workhorse.
Gemma 2 9B
9B Gemma 2 — single-GPU local target.
Gemma 2 2B
Tiny 2B Gemma 2 — laptop / mobile.
Gemma 3 4B
4B Gemma 3 — laptop multimodal.
Gemma 3 1B
1B Gemma 3 — edge / mobile.
Gemma 4 12B It
Gemma 2B It
Gemma 3 1B Pt
Gemma 3 270M It
Gemma 3 27B Pt
Gemma 4 E2B-it
Gemma 4 E4B-it
Google: Gemma 3n 4B
Gemma 3n E4B-it is optimized for efficient execution on mobile and low-resour...
Google: Gemma 4 26B A4B
Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model fro...
Google: Gemma 4 31B
Gemma 4 31B Instruct is Google DeepMind's 30.7B dense multimodal model suppor...
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)
|
32 GB | — | — | Compare → | |
|
FP8
FP8 — 8-bit float (Hopper / Blackwell)
|
16 GB | $0.10/hr | — | Compare → | |
|
INT4
INT4 — 4-bit integer (~4× VRAM saving)
|
8 GB | — | — | Compare → |
What it costs per month across providers.
Estimate your monthly bill for Google: Gemma 3 12B across every host that publishes per-token pricing. Slide your token volumes; the chart + table re-rank cheapest-first.
Cheapest provider on the left.
Total monthly cost — input + output tokens combined.
Bill breakdown.
Rent the GPU instead of paying per token.
For an open-weights model like Google: Gemma 3 12B, you can rent a GPU and serve inference yourself. The math: cheapest GPU rental × 730 hours/month + your electricity rate × power draw.
Assumes the GPU runs 24/7 at ~85% utilisation. If your traffic is bursty, you'll pay less for the API and probably more for the GPU (idle hours still cost rental). The breakeven analysis lives on the Self-host vs API breakeven tool.