Gemma 3 27B.
Google's open-weight multimodal LLM — efficient and license-permissive.
1× Nvidia RTX 4090.
Most-aggressive quantisation we have a working recommendation for. Lower precision = less VRAM = cheaper hardware, at a small accuracy cost.
Cheapest hosted endpoints.
| Provider | Access | $/M in | $/M out | |
|---|---|---|---|---|
| OpenRouter | api aggregator | $0.08 | $0.16 | Launch ↗ |
What it's best at.
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 | 44.9 | 782 ms | 2648 ms |
Smaller models distilled from Gemma 3 27B.
Lightweight student models trained to mimic Gemma 3 27B's outputs.
Variants in the Gemma family.
Workloads.
Frequently asked.
How do I run Gemma 3 27B?
Where can I access Gemma 3 27B?
How much does it cost to run Gemma 3 27B?
Is Gemma 3 27B open-source or proprietary?
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)
|
48 GB | $0.28/hr | — | Compare → | |
|
FP8
FP8 — 8-bit float (Hopper / Blackwell)
|
40 GB | — | — | Compare → | |
|
INT4
INT4 — 4-bit integer (~4× VRAM saving)
|
Snug fit, recommended for hobbyist self-hosting
|
24 GB | — | — | Compare → |
What it costs per month across providers.
Estimate your monthly bill for Gemma 3 27B 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.
| Provider | Monthly total | |
|---|---|---|
| $1.12 | Sign up ↗ |
Rent the GPU instead of paying per token.
For an open-weights model like Gemma 3 27B, 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.
What it's best at.
Scores normalised against benchmark ceilings (100 = perfect). Coloured by tier — coral 80+ frontier, lavender 65+ strong, sage 50+ solid, slate below.
Published scores.
| Benchmark | Score | Source |
|---|---|---|
| GPQA | 42.4 | official ↗ |
| MATH | 89.0 | official ↗ |
| MMLU | 76.2 | official ↗ |
| MMMU | 64.9 | official ↗ |
| MMLU-Pro | 67.5 | official ↗ |
| HumanEval | 81.0 | official ↗ |
Independent rankings.
About Gemma 3 27B.
Gemma 3 27B is Google's open-weight model in the Gemma family — distilled from the Gemini architecture, released under the Gemma Terms of Use (commercial use allowed with attribution). The 3 series introduced native multimodal support (vision input) and a 128K context window — both firsts for an open-weight Google model. Runs comfortably on a single H100 (FP16) or RTX 5090 (INT4). Strong on multilingual tasks (140+ languages) and competitive with Llama 3.3 70B at half the parameter count.