Gemini 3.1 Flash Lite Preview
Gemini 3.1 Flash Lite Preview is the efficiency-focused model in the Gemini 3.1 generation for budget-constrained, high-volume workloads, with notable gains in translation, data extraction, and code completion over Gemini 2.5 Flash Lite and four configurable thinking levels.
import { streamText } from 'ai'
const result = streamText({ model: 'google/gemini-3.1-flash-lite-preview', prompt: 'Why is the sky blue?'})Playground
Try out Gemini 3.1 Flash Lite Preview by Google. Usage is billed to your team at API rates. Free users (those who haven't made a payment) get $5 of credits every 30 days.
Providers
Route requests across multiple providers. Copy a provider slug to set your preference. Visit the docs for more info. Using a provider means you agree to their terms, listed under Legal.
P50 throughput on live AI Gateway traffic, in tokens per second (TPS). Visit the docs for more info.
P50 time to first token (TTFT) on live AI Gateway traffic, in milliseconds. View the docs for more info.
Direct request success rate on AI Gateway and per-provider. Visit the docs for more info.
More models by Google
| Model |
|---|
| 1M | 2.9s | 323tps | $1.50/M | $9.00/M | Read:$0.15/M Write:— | $14.00/K + input costs | — |   | 05/19/2026 | ||||
| 1M | 3.5s | 131tps | $2.00/M | $12.00/M | Read: $0.2/M Write: — | $14.00/K + input costs | — | | 02/19/2026 | ||||
| 1M | 0.7s | 179tps | $0.50/M | $3.00/M | Read: $0.05/M Write: — | $14.00/K + input costs | — | | 12/17/2025 | ||||
| 1M | 0.5s | 248tps | $0.10/M | $0.40/M | Read:$0.01/M Write:— | $35.00/K + input costs | — | | 06/17/2025 | ||||
| 1M | 0.3s | 192tps | $0.30/M | $2.50/M | Read:$0.03/M Write:— | $35.00/K + input costs | — | | 03/20/2025 | ||||
| 1M | 2.0s | 123tps | $1.25/M | $10.00/M | Read: $0.13/M Write: — | $35.00/K + input costs | — | | 03/20/2025 |
About Gemini 3.1 Flash Lite Preview
Gemini 3.1 Flash Lite Preview is Google's most cost-efficient model in the 3.1 generation, designed explicitly for high-volume agentic tasks, data extraction pipelines, and latency-sensitive applications where budget is the primary constraint. This model outperforms Gemini 2.5 Flash Lite on overall quality, with the most pronounced improvements in translation, data extraction, and code completion, three task categories that commonly drive the highest request volumes in production.
The four-level thinking configuration (minimal, low, medium, high) is a notable engineering affordance. It allows a single model deployment to serve heterogeneous workloads without switching models: a bulk extraction job might run at minimal thinking to minimize latency and cost, while an edge-case translation that requires cultural nuance detection runs at medium. For teams running large-scale pipelines, content localization, automated data cleaning, code completion at IDE scale, or classification across millions of documents, Gemini 3.1 Flash Lite Preview provides the quality improvements of the 3.1 generation without the cost profile of the Pro or standard Flash tiers. Its position in the lineup is defined by throughput economics rather than maximum capability.
What To Consider When Choosing a Provider
- Configuration: Gemini 3.1 Flash Lite Preview supports four thinking levels,
minimal,low,medium, andhigh, giving you fine-grained control over the reasoning-to-cost tradeoff across different task types within the same deployment. - Zero Data Retention: AI Gateway supports Zero Data Retention for this model via direct gateway requests (BYOK is not included). To configure this, check the documentation.
- Authentication: AI Gateway authenticates requests using an API key or OIDC token. You do not need to manage provider credentials directly.
When to Use Gemini 3.1 Flash Lite Preview
Best For
- High-volume data extraction: Pipelines processing millions of structured or semi-structured inputs
- Bulk translation workloads: Per-character cost directly determines operational feasibility
- Code completion and review: Inline suggestions and automated review at IDE or CI/CD pipeline scale
- Parallel sub-agent systems: Agentic architectures where aggregate token cost is a binding constraint
- Latency-first applications: Task complexity is moderate to low and response speed is the primary criterion
Consider Alternatives When
- Lite-tier quality insufficient: Your task quality cannot tolerate the tradeoffs (consider
google/gemini-3.1-flash-image-previeworgoogle/gemini-3-flashfor higher quality) - Multi-step reasoning: Your workflow involves complex documents or images (consider
google/gemini-3.1-pro-previeworgoogle/gemini-3-pro-preview) - Native image generation: You need image output alongside text (consider
google/gemini-3.1-flash-image-preview) - Maximum reasoning depth: The task requires deepest reasoning without cost constraints (consider
google/gemini-3.1-pro-preview)
Conclusion
Gemini 3.1 Flash Lite Preview brings the quality advances of the 3.1 generation to the highest-volume, most cost-sensitive segment of the model market. With four thinking levels and documented improvements in the task categories that drive the most tokens in production, it's the natural choice for teams optimizing for scale economics rather than peak capability.
Frequently Asked Questions
What are the four thinking levels and what do they mean for cost and quality?
minimal,low,medium, andhigh. Lower levels reduce the amount of reasoning compute applied before generating a response, which decreases latency and token consumption but may reduce quality on complex tasks.highapplies the most reasoning, similar to configuring a reasoning model for thorough inference.Which task categories saw the most improvement over Gemini 2.5 Flash Lite?
Translation, data extraction, and code completion saw the largest improvements over Gemini 2.5 Flash Lite. These are the high-volume task categories where the efficiency gains of the 3.1 generation have the most practical impact.
Is this model suitable for agentic multi-agent architectures?
Yes. High-volume agentic tasks are a primary target. The model's low cost and configurable thinking levels make it appropriate for sub-agents in hierarchical agent systems.
Can I mix thinking levels in the same application?
Yes. You set
thinkingLevelper request inproviderOptions.google.thinkingConfig, so different request types within the same application can use different levels without any architectural changes.Does Gemini 3.1 Flash Lite Preview support streaming?
Yes. Use
streamTextfrom the AI SDK withmodel: 'google/gemini-3.1-flash-lite-preview'to stream responses.How does Gemini 3.1 Flash Lite Preview differ from Gemini 3 Flash?
Gemini 3 Flash prioritizes pro-grade reasoning at flash speed and is positioned as the standard speed/quality balance point. Flash Lite is specifically optimized for maximum cost efficiency and high volume throughput, trading some capability headroom for a lower price point.
Is
includeThoughtssupported on this model?Yes. Set
includeThoughts: trueinproviderOptions.google.thinkingConfigto stream the model's reasoning tokens alongside the generated response.What's the recommended thinking level for bulk translation tasks?
Start with
loworminimalfor straightforward translation tasks where throughput is the primary concern. Increase tomediumfor content requiring cultural nuance or domain-specific accuracy.