Google 推出的 Gemma 3n 是一款专注于超自然语音合成、实时音频理解与低延迟语音交互的大模型，具备 32K 上下文窗口、150 Token/秒生成速度与 52/100 综合智能指数（免费开源权重）。
Technical Specifications
- Provider
- Category
- open-source
- Release Date
- 2025-06
- Batch API Discount
- Standard rate
- Consumer Subscription
- Free / Self-Hosted
- Reasoning Effort Control
- Fixed
- Structured Outputs
- Yes
- Output Modalities
- text
- Official Documentation
- Visit Official Docs ↗
Data Privacy, Licensing & Deployment Governance
Official Provider Release Intelligence, API Quotas & Agentic Specs
- Official API Model ID (SDK String)
- models/gemma-3n
- Supported SDK & API Protocol
- Google Gemini API (generateContent) + OpenAI-Compatible SDK
- Lifecycle & Deprecation Status
- Active GA (Guaranteed support through 2027+)
- Production Rate Limits (RPM & TPM)
- Unlimited (Hardware-bound on self-hosted GPU) / 10,000 RPM Cloud · Unlimited local throughput / 10M+ TPM Managed Cloud
- Long-Context Pricing Tier (>128K / >200K)
- Free across full context window (only KV-cache VRAM scales)
- Prompt Cache Write & TTL Policy
- Automatic prefix KV-caching in vLLM / SGLang (0 cost)
- Computer Use & Browser GUI Agent
- Supported (Screen coordinate grounding, browser & desktop GUI action loops)
- MCP & Parallel Tool Calling
- Native MCP Client/Server ready · Parallel function calling · 100% Strict JSON Schema
- Web Search Grounding & Code Sandbox
- Built-in Google Search Grounding + Managed Python Code Execution Sandbox
- Fine-Tuning & Distillation Support
- Full Weights SFT, LoRA, QLoRA, DPO & GRPO RL (Unsloth / Axolotl / Llama-Factory)
- Multimodal Tokenization Billing
- Text BPE (~4 chars/token) · Vision (~258–765 tokens per image tile) · Audio (~25–32 tokens/sec)
- Weights Precision & Quantization
- BF16 / FP16, FP8 Dynamic, INT4 AWQ / GPTQ, GGUF (Q4_K_M / Q8_0)
- Minimum GPU VRAM Requirement
- 2 GB VRAM (4-bit Quantized) · 4 GB VRAM (Native BF16)
- Supported Inference Engines
- vLLM, SGLang, Ollama, llama.cpp, TensorRT-LLM, HuggingFace TGI
- System Card & Safety Framework
- Google DeepMind Frontier Safety Framework Verified
- Enterprise Compliance (SOC2 / HIPAA / ZDR)
- 100% Air-Gapped On-Premises / Private VPC · Full HIPAA, GDPR & SOC2 Control
Gemma 3n Cost, Context, Hardware & API Integration
Open-weights or free-tier model: $0 API license cost when self-hosted.
Same Monthly Workload Cost Across Popular Reference Models
Live Prompt Paste Tokenizer & Cost Counter
Paste real system instructions, JSON schemas, or code to measure BPE token counts across tokenizers.
Hybrid Prompt Routing Cascade Simulator
Route routine queries to a fast lightweight model ($0.15/1M) and escalate only complex reasoning turns to Gemma 3n:
Benchmark Scores & 6-Axis Capability Radar
- MMLU-Pro
- 64%
- HumanEval
- 71%
- Chatbot Arena Elo
- 1220pts
Strengths
- 提供 32K 上下文窗口与最高 4K 输出容量，核心基准表现优异（MMLU-Pro 64% · Arena Elo 1220）
- 高达 150 Token/秒的生成吞吐量，首字延迟 (TTFT) 仅约 220 毫秒
- 开放模型权重 (2B / 5B)，支持通过 vLLM、SGLang 或 Ollama 在私有 GPU 集群上本地部署与微调
Limitations & Trade-offs
- 在本地以满血精度运行需要较高的 GPU 显存 (VRAM) 与运维基础设施
- 在超长链路的前沿科研推导任务上，深度略逊于顶配重型旗舰推理模型
Editorial Assessment
Gemma 3n 凭借 52/100 的智能得分、32K 上下文容量与 150 Token/秒的响应速度，在 Google 模型矩阵中提供了极佳的性能与成本平衡。
Best for: 专注于超自然语音合成、实时音频理解与低延迟语音交互的任务、企业级 RAG 检索增强生成、自动化智能体工作流与高可靠 API 集成。
Compare Gemma 3n
Frequently Asked Questions about Gemma 3n
How much does the Gemma 3n API cost per 1M tokens?+
Gemma 3n by Google is priced at $0 per 1M input tokens and $0 per 1M output tokens. With prompt caching enabled, cached input tokens cost $0 per 1M tokens, and asynchronous Batch API requests receive a 0% discount.
Should I use a $20/month subscription or the Gemma 3n pay-as-you-go API?+
A flat $20/month consumer subscription is predictable for daily interactive chat and manual coding. However, if you are building software, running automated workflows, or sending low-to-moderate volume with prompt caching enabled, the pay-as-you-go API is typically far more cost-effective and offers higher rate limits without manual UI caps.
How many words and pages fit in Gemma 3n's context window?+
Gemma 3n features a 32K tokens context window, which equals approximately 24,000 English words or ~64 standard single-spaced pages in a single prompt. Its maximum output generation per turn is 4,096 tokens.
Can I run Gemma 3n locally on my own GPU (Open Weights)?+
Yes, Gemma 3n (2B / 5B) releases open weights that can be self-hosted locally or on private cloud GPUs using vLLM, Ollama, SGLang, or llama.cpp.
Does Google train on data sent to the Gemma 3n API, and is commercial use allowed?+
100% Isolated when self-hosted on private GPUs or VPC. License tier: Open-Weights Research & Commercial License. Full commercial ownership of generated outputs. Supported deployment channels include: Self-Hosted GPU (vLLM / Ollama), Private VPC, or Managed Serverless API.
How does Gemma 3n perform on coding and reasoning benchmarks?+
Gemma 3n achieves a composite Intelligence Index of 52/100, scoring N/A% on SWE-bench Verified, N/A% on GPQA Diamond, 64% on MMLU-Pro, and 1220 pts on Chatbot Arena Elo.
What is Gemma 3n best used for?+
Gemma 3n is best suited for: 专注于超自然语音合成、实时音频理解与低延迟语音交互的任务、企业级 RAG 检索增强生成、自动化智能体工作流与高可靠 API 集成。. Assessment: Gemma 3n 凭借 52/100 的智能得分、32K 上下文容量与 150 Token/秒的响应速度，在 Google 模型矩阵中提供了极佳的性能与成本平衡。
Google Family Lineage, Routing Tiers & Upgrade Path
Step down to a lighter sibling for high-volume routing or step up to Google's flagship reasoning tier:
Gemini 4 Argon
Gemini 3.8 Flash
Gemini 3.7 Flash
Gemini 3.6 Flash
Gemini 3.5 Flash-Lite
Gemini Omni Flash
AI Glossary & Architecture Concepts
View full glossary →Key machine learning concepts and benchmarks relevant to Gemma 3n:
The basic unit of text that a language model processes, typically representing about 3 to 4 English characters or roughly 0.75 words. For example, the sentence "Artificial intelligence is fast" breaks down into about 5 tokens.
A standardized evaluation suite used to objectively measure and rank an AI model's capabilities across specific tasks — such as SWE-bench Verified for software engineering, AIME for competitive math, or MMLU for general knowledge.
The maximum number of tokens a model can consider at once — both the input it reads and the output it writes. A 1,000,000-token context window can hold roughly 750,000 English words in a single request.
The ability of a language model to recognize when an external capability is needed and output structured JSON arguments to invoke external APIs, databases, or terminal commands. The runtime executes the tool and passes results back to the model.
An open standard created by Anthropic that allows AI applications and agents to connect seamlessly to local filesystems, development environments, and third-party APIs via standardized server interfaces.
A multimodal neural network engineered to jointly process, reason over, and synthesize both visual content (images, diagrams, UI screenshots, video) and natural language text within a single framework.
Benchmark scores, token prices, and rate limits evolve rapidly as AI labs release model updates. Always verify current specifications and pricing on the official provider documentation.