Google Gemma 3 27B Instruct
Google's 27B open-weight multimodal model with vision, tool use, and 140+ language support — efficient enough for consumer hardware.
Get API key- Provider
- Google DeepMind
- Price
- $0.12 in · $0.20 out / 1M
- Context window
- 198K tokens
- Released
- March 12, 2025
- License
- Gemma license (open weights, usage restrictions)
What is Google Gemma 3 27B Instruct?
Google Gemma 3 27B Instruct is a 27-billion-parameter multimodal language model from Google DeepMind, released in March 2025. It accepts text and image inputs, generates text outputs, supports tool use and structured outputs, and ships with open weights under the Gemma license.
Use it privately on Venice
On Venice, Gemma 3 27B runs under a private, zero-retention tier — your prompts are not stored or used for training. You get the full open-weight model with vision, tool use, and web search capabilities without surrendering data sovereignty to Big-Tech inference clouds.
What can it do?
- •Open weights under the Gemma license, enabling local deployment, fine-tuning, and full model sovereignty.
- •Native vision support with multiple image inputs and structured JSON output for building multimodal applications.
- •Tool use and function calling plus web search integration, allowing agentic workflows and grounded responses.
- •Extremely cost-efficient on Venice — among the cheapest vision-capable models at $0.12 in / $0.20 out per million tokens.
- •Multilingual coverage across 140+ languages and a lightweight footprint that runs well on consumer GPUs.
- •Not uncensored — it carries standard Gemma safety alignments and may refuse sensitive or harmful requests.
- •Reasoning and coding benchmarks trail larger proprietary models and top-tier open rivals such as DeepSeek V3.2.
- •The 27B parameter count limits performance on highly complex agentic pipelines and advanced software engineering tasks compared to 70B+ or frontier models.
- •The Gemma license includes usage restrictions that may limit commercial redistribution or modification compared to permissive open-source licenses.
Google Gemma 3 27B Instruct capabilities
- Tool use / function calling
- Vision (image input)
- Reasoning
- Web search
- Code-optimized
- Structured output (JSON schema)
- Audio input
- Video input
- Multiple image inputs
- Log probabilities
How to use it via API
Venice exposes an OpenAI-compatible API. Swap your base URL and call google-gemma-3-27b-it.
curl https://api.venice.ai/api/v1/chat/completions \
-H "Authorization: Bearer $VENICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "google-gemma-3-27b-it",
"messages": [{ "role": "user", "content": "Explain quantum tunneling simply." }]
}'Specifications
Pricing
Billed per token on Venice: $0.12 per 1M input tokens and $0.20 per 1M output tokens.
New Venice accounts include a free daily allowance and 500 welcome credits — no credit card required.
Google Gemma 3 27B Instruct vs alternatives
| Model | Context window | Open weights | Price (Venice) |
|---|---|---|---|
| Google Gemma 3 27B Instruct | 198K tokens | Yes | $0.12 in · $0.20 out / 1M |
| Google Gemma 4 31B Instruct | 256K tokens | Yes | $0.12 in · $0.36 out / 1M |
| DeepSeek V3.2 | 160K tokens | Yes | $0.33 in · $0.48 out / 1M |
| Claude Sonnet 4.6 | 1M tokens | No | $3.60 in · $18 out / 1M |
A lightweight, open-weight multimodal workhorse with vision, tool use, and web search at a fraction of frontier pricing.
What is it good for?
- •Multilingual chatbots and customer support across 140+ languages.
- •Vision-enabled applications that analyze images, charts, or UI mockups alongside text prompts.
- •Lightweight agentic workflows using tool calling and structured JSON output.
- •Local or private-cloud deployment for organizations that require data sovereignty.
- •Cost-efficient content generation, summarization, and question answering at high volume.
Prompting tips
- •Use explicit JSON schemas when requesting structured output — the model follows them reliably.
- •For vision tasks, describe exactly what you want extracted from the image; it handles multiple image inputs in one prompt.
- •Enable web search when questions need current facts beyond the training cutoff.
- •Break complex coding or reasoning tasks into smaller single-turn prompts rather than long multi-step chains.
Version history
CurrentCurrent release — adds multimodality, tool use, and expanded language support.
Frequently asked questions
Google Gemma 3 27B Instruct is a 27-billion-parameter multimodal language model from Google DeepMind, released in March 2025. It processes text and image inputs, generates text outputs, and supports tool use, function calling, and structured JSON output under open weights.
Venice charges $0.12 per 1 million input tokens and $0.20 per 1 million output tokens. There is no subscription required; you pay only for the tokens you use.
It is an open-weights model released by Google under the Gemma license, which permits commercial use with certain restrictions. It is not OSI-approved open source, but you can download the weights from Hugging Face and run them locally or on Venice.
Yes. On Venice it supports function calling, tool use, vision (image input), multiple image inputs per prompt, structured JSON output, and web search.
Choose Gemma 3 27B if you want the lowest cost for a vision-capable open model and native multilingual support. Choose DeepSeek V3.2 if you need stronger coding and reasoning performance and are willing to pay a higher per-token rate.
No. It is not uncensored; it retains standard Gemma safety alignments and may refuse certain sensitive or harmful requests. For fully uncensored open-weight alternatives, browse Venice's model directory.
Yes. The open weights are available on Hugging Face and can be run with Ollama, llama.cpp, or vLLM. Venice also hosts it privately if you prefer not to manage infrastructure.
It supports more than 140 languages natively, making it suitable for global applications and multilingual content generation.
Related models
Run Google Gemma 3 27B Instruct privately.
No prompt logging. No data used for training. Free to start — no credit card.
