Gemma 3.2
Google's most capable open multimodal model, combining image and text understanding with support for 140+ languages and a massive 128K context window.
Google 
DeepInfra Open multimodal model with 128K context
Gemma 3 27B is an open-weight multimodal language model developed by Google DeepMind as the flagship model in the Gemma 3 family. It accepts both image and text inputs and generates text outputs, supporting over 140 languages and a context window of 128,000 tokens — sixteen times larger than the previous Gemma 2 generation. The model is built on the same research foundation as Google's Gemini models and was released in March 2025.
Gemma 3 27B is designed to run in resource-constrained environments, including on a single consumer GPU with 24GB of VRAM, as well as on laptops, desktops, and cloud infrastructure. It is well-suited for tasks such as visual question answering, document analysis, multilingual text generation, summarization, coding assistance, and logical reasoning. Its combination of multimodal input support, large context handling, and open-weight availability makes it a practical choice for developers building applications that require flexible deployment options.
What Gemma 3.2 supports
Multimodal Input
Processes both images and text as input in a single request, enabling tasks like visual question answering, image description, and document analysis.
128K Context Window
Handles up to 128,000 tokens of input per request, allowing analysis of long documents, large codebases, and extended multi-turn conversations.
Multilingual Generation
Generates and understands text in 140+ languages, making it suitable for globally-facing applications and cross-language tasks.
Reasoning & Analysis
Performs multi-step logical reasoning, summarization, and question answering across complex inputs including code and structured documents.
Code Generation
Generates, explains, and debugs code across common programming languages as part of its general text generation capabilities.
Flexible Deployment
Runs locally on a single GPU with 24GB VRAM (e.g., RTX 3090) as well as on cloud infrastructure, with open weights available for download.
Ready to build with Gemma 3.2?
Get Started FreeBenchmark scores
Scores represent accuracy — the percentage of questions answered correctly on each test.
| Benchmark | What it tests | Score |
|---|---|---|
| MMLU-Pro | Expert knowledge across 14 academic disciplines | 66.9% |
| GPQA Diamond | PhD-level science questions (biology, physics, chemistry) | 42.8% |
| MATH-500 | Undergraduate and competition-level math problems | 88.3% |
| AIME 2024 | American math olympiad problems | 25.3% |
| LiveCodeBench | Real-world coding tasks from recent competitions | 13.7% |
| HLE | Questions that challenge frontier models across many domains | 4.7% |
| SciCode | Scientific research coding and numerical methods | 21.2% |
Common questions about Gemma 3.2
What is the context window size for Gemma 3 27B?
Gemma 3 27B supports a context window of 128,000 tokens, which is sixteen times larger than the previous Gemma 2 generation.
Does Gemma 3 27B support image inputs?
Yes. Gemma 3 27B is a multimodal model that accepts both image and text as inputs, enabling tasks such as visual question answering, image description, and document analysis.
What languages does Gemma 3 27B support?
The model supports over 140 languages for both input understanding and text generation.
What is the training data cutoff for Gemma 3 27B?
Based on the available metadata, the model's training date is listed as March 2025. For precise knowledge cutoff details, refer to the official Gemma 3 Technical Report.
Can Gemma 3 27B be run locally?
Yes. The model is open-weight and can be deployed locally on a single GPU with 24GB of VRAM, such as an NVIDIA RTX 3090, as well as on laptops, desktops, or cloud infrastructure.
Is Gemma 3 27B free to use?
Gemma 3 27B is an open-weight model, meaning the weights are publicly available. Usage costs on MindStudio depend on the underlying inference provider (DeepInfra in this case); consult MindStudio's pricing page for current rates.
Documentation & links
Parameters & options
Explore similar models
Google Gemini 3.1 Pro
Google's frontier reasoning model delivering enhanced software engineering, agentic reliability, and multimodal intelligence across a 1M-token context window.
Google Gemini 3
The best model in the world for multimodal understanding, and our most powerful agentic and vibe-coding model yet, delivering richer visuals and deeper interactivity, all built on a foundation of state-of-the-art reasoning.
Google Gemini 2.5 Pro
State-of-the-art thinking model, capable of reasoning over complex problems in code, math, and STEM, as well as analyzing large datasets, codebases, and documents using long context.
OpenAI GPT‑5.2 Pro
The most capable model series yet for professional knowledge work.
Anthropic Claude 4.6 Sonnet
Anthropic's most capable Sonnet model yet, delivering near-Opus-level intelligence for coding, agents, and computer use with a 1M token context window.
Mistral Mistral Medium 3
A versatile AI model designed for professional applications with particular strengths in coding and multimodal understanding. The model can be deployed across various environments, making it ideal for enhancing customer service, personalizing business processes, and analyzing complex datasets across industries.
Start building with Gemma 3.2
No API keys required. Create AI-powered workflows with Gemma 3.2 in minutes — free.