Gemma 3 12B
Gemma 3 introduces multimodality, supporting vision-language input and text outputs.
text + image → text · made by Google
Sold by 10 ways
| Seller | Lane | Rate |
|---|---|---|
| AWS Bedrock cloud | standard | $0.09per Mtok in$0.29per Mtok outmodels.dev · read 2026-09-16 |
| Crusoe cloud | standard | $0.1per Mtok in$0.1per Mtok outraw.githubusercontent.com · read 2026-09-16 |
| Databricks cloud | standard | $0.15per Mtok in$0.5per Mtok out$0.15per Mtok cached$0.15per Mtok cache writeraw.githubusercontent.com · read 2026-09-16 |
| DeepInfra aggregator | standard | $0.05per Mtok in$0.15per Mtok outapi.deepinfra.com · read 2026-08-25 |
| Hugging Face Inference Providers 2 ways | $0.05per Mtok in$0.15per Mtok out | |
| Hugging Face Inference Providers aggregator | standard | $0.05per Mtok in$0.15per Mtok outmodels.dev · read 2026-09-16 |
| Hugging Face Inference Providers aggregator | deepinfra | $0.05per Mtok in$0.15per Mtok outrouter.huggingface.co · read 2026-08-25 |
| Nous Research aggregator | standard | $0.05per Mtok in$0.15per Mtok outinference-api.nousresearch.com · read 2026-09-12 |
| Novita AI aggregator | standard | $0.05per Mtok in$0.1per Mtok outapi.novita.ai · read 2026-08-25 |
| OpenRouter aggregator | standard | $0.05per Mtok in$0.15per Mtok outopenrouter.ai · read 2026-08-24 |
| ElectronHub aggregator | standard | $0.35per Mtok in$0.56per Mtok outapi.electronhub.ai · read 2026-09-16 |
Measured 12 standings
| Place | Board | Metric | Score |
|---|---|---|---|
| 7thof 105 | Vectara Hallucination Leaderboard | Hallucination Rate | 4.4 |
| 122ndof 123 | Lmca — Epoch AI | Score | 4.46 |
| 142ndof 143 | Artificial Analysis · Agentic Index | Index | 0.1 |
| 149thof 161 | Dtbench — Epoch AI | Accuracy | 48.8 |
| 164thof 177 | Critpt — Epoch AI | Accuracy | 0 |
| 164thof 168 | Scicode — Epoch AI | Score | 17.361 |
| 181stof 184 | Artificial Analysis · Coding Index | Index | 5.8 |
| 214thof 222 | Chess puzzles — Epoch AI | mean_score | 0 |
| 220thof 402 | LMArena · Text | Score (Elo) | 1342 |
| 232ndof 290 | OTIS Mock AIME 2024-2025 | mean_score | 16.667 |
| 268thof 313 | GPQA Diamond | mean_score | 39.457 |
| 561stof 656 | Epoch capabilities index — Epoch AI | ECI Score | 123.48 |
About
Gemma 3 12B — a text model from Google, sold by 9 companies from $0.05 in and $0.15 out per million tokens, placed 7th of 105 on Vectara Hallucination Leaderboard.
It takes text and images and returns text, with a context window of 131,072 tokens. It was published in December 2024, trained on material up to August 2024. The catalogue files it under reasoning. Nine companies sell it. The cheapest is $0.05 in and $0.15 out per million tokens at DeepInfra. Beside the standard rate there is a separately routed lane. It has been measured on 12 boards, and stands best at 7th of 105 on Vectara Hallucination Leaderboard.
Every current figure
- Maker
- Register
- model
- Takes
- text + image
- Returns
- text
- Context
- 131,072 tokens
- Published
- December 2024
- Knowledge to
- August 2024
- Parameters
- 12 billion
- Licence
- not read
- Sellers
- 9
- Maker's own price
- not read
- Price
- $0.05 in and $0.15 out per million tokens — DeepInfra
- Boards
- 12
- Best place
- 7th of 105 — Vectara Hallucination Leaderboard
Known as 1 name
google/gemma-3-12b-it