The request names gemma-3-12b and MT6-Gen3, which is not identified in any source. The closest match is Google's Gemma 3 12B instruct model, which is offered through hosted APIs such as Replicate 110 and OpenRouter 413, and through Google's own Gemini Developer API page for Gemma open models 716. Which platform or deployment you want (hosted API, self-hosted, or on-device) would decide the best choice, and it is not stated.
A question back: Which platform do you want to run MT6-Gen3_gemma-3-12B on: a hosted API, your own hardware, or on-device? And what is MT6-Gen3, a fine-tune or a specific build?
- 1
Replicatereplicate.comverified and signed
Runs open-source models such as google-deepmind/gemma-3-12b-it through a cloud API, where Gemma is described as a lightweight open model family from Google 10.
for agents: google-deepmind/gemma-3-12b-it
- 2
OpenRouteropenrouter.aiverified and signed
Provides a unified API across 500+ models, with tool-calling and text+image-to-text capabilities listed for the models it routes 13.
for agents: tool-calling · structured-outputs · text+image->text
- 3
Gemini Developer API (Google AI for Developers)google.comverified and signed
Offers Google's Gemini API and Google AI Studio, and Gemma open models that can be tailored and run on-device with Google AI Edge 716.
official:SDK · @google/gemini-clirepository · sec-geminiskill · gemini-agents-apiskill · gemini-apiskill · gemini-interactions-api
for agents: gemini-api · gemini-omni-flash-api · gemini-live-api
Sources
- 1Replicate - Run AI with an APIreplicate.com · read 2026-10-29
- 4OpenRouteropenrouter.ai · read 2026-10-29
- 7Gemini Developer API | Gemma open models | Google AI for Developersai.google.dev · read 2026-10-29
- 10google-deepmind/gemma-3-12b-it: declares "google-deepmind/gemma-3-12b-it" (the record we read and signed)unfragile.ai · read 2026-10-24
- 13Google: Gemma 3 12B: declares "tool-calling" (the record we read and signed)unfragile.ai · read 2026-10-24
- 16gemini-api: declares "gemini-api" (the record we read and signed)unfragile.ai · read 2026-10-24
For agents: the 10 actions behind this answer
google-deepmind/gemma-3-12b-it · google-deepmind/gemma-3-12b-it verified and signed
Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models.
nano-banana-pro-openrouter · nano-banana-pro-openrouter verified and signed
Generate or edit images via OpenRouter with the Gemini 3 Pro Image model. Use for prompt-only image generation, image edits, and multi-image compositing; supports 1K/2K/4K output.
google-deepmind/gemma-3-27b-it · google-deepmind/gemma-3-27b-it verified and signed
Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models.
google-deepmind/gemma-3-4b-it · google-deepmind/gemma-3-4b-it verified and signed
Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models.
google-deepmind/shieldgemma-2-4b-it · google-deepmind/shieldgemma-2-4b-it verified and signed
ShieldGemma 2 is a model trained on Gemma 3's 4B IT checkpoint for image safety classification across key categories that takes in images and outputs safety labels per policy.
prunaai/gemma-4-26b-a4b-fast · prunaai/gemma-4-26b-a4b-fast verified and signed
This is a version of the MoE Gemma 4 26B optimised by Pruna AI.
google-deepmind/gemma-7b-it · google-deepmind/gemma-7b-it verified and signed
7B instruct version of Google’s Gemma model
workers-ai-search-model · Cloudflare API verified and signed
Model Search
lucataco/gemma-4-31b-it · lucataco/gemma-4-31b-it verified and signed
Gemma 4 31B Instruct - Google open-weight VLM (image + text in, text out)
google/gemini-3-flash · google/gemini-3-flash verified and signed
Google's most intelligent model built for speed with frontier intelligence, superior search, and grounding
Labels are the instruments' verdict and are never changed by ranking. Agents and robots get this same answer as JSON at /api/v1/find?q=…&answer=1 and through the MCP server.