Gemma 3 12B IT
google/gemma-3-12b-it
spread (output, p90 ÷ p10)
1.0×
$0.100 – $0.290 · 4 providers · max÷min 2.9×
Cheapest / median / spread use providers that serve the model on their own capacity: first-party lab, hyperscaler, inference host.
blended = 90.9% input × [(1 − 70% cache) × input + 70% cache × cache-read] + 9.1% × output + cache writes (30% of uncached input) where a write price is published
A provider that publishes no cache price is blended with cache share 0 (all input at the input price) and labelled “no cache price published” — never penalised against identical in/out prices.
| Provider | Input | Output | Cache rd | Blended |
|---|---|---|---|---|
Helicone router / aggregator | $0.0500 | $0.100 | — | $0.0545no cache price published |
NovitaAI inference host | $0.0500 | $0.100 | — | $0.0545no cache price published |
Hugging Face inference host | $0.0500 | $0.150 | — | $0.0591no cache price published |
Kilo Gateway router / aggregator | $0.0500 | $0.150 | — | $0.0591no cache price published |
OpenRouter router / aggregator | $0.0500 | $0.150 | — | $0.0591no cache price published |
Deep Infra inference host | $0.0500 | $0.150 | — | $0.0591no cache price published |
Eden AI router / aggregator | $0.0500 | $0.150 | — | $0.0591no cache price published |
NanoGPT router / aggregator | $0.272 | $0.272 | $0.136 | $0.185 |
Merge Gateway router / aggregator | $0.0900 | $0.290 | — | $0.108no cache price published |
Amazon Bedrock hyperscaler | $0.0900 | $0.290 | — | $0.108no cache price published |
Free tiers (rate-limited) · 1
No payment required; rate limits and quotas apply (not tracked yet). Excluded from cheapest, median and spread.
- Nvidia google/gemma-3-12b-itNVIDIA developer account + API keyfree tier
Data: models.dev (MIT), snapshot 2026-09-20. Prices are vendor-published; verify before purchase. Methodology.