Repository Public
1Description: Learn how to optimize latency with the Priority inference tier23The Gemini Priority API is a premium inference tier designed for4business-critical workloads that require lower latency and the highest5reliability at a premium price point. Priority tier traffic is prioritized above6standard API and Flex tier traffic.78Priority inference is available to [Tier 2 \& Tier 3](https://ai.google.dev/gemini-api/docs/billing#about-billing) users across the GenerateContent API9and Interactions API endpoints.1011## How to use Priority1213To use the Priority tier, set the `service_tier` field in the request body to14`priority`. The default tier is standard if the field is omitted.1516### Python1718 from google import genai1920 client = genai.Client()2122 try:23 response = client.models.generate_content(24 model="gemini-3.8-flash",25 contents="Triage this critical customer support ticket immediately.",26 config={"service_tier": "priority"},27 )2829 # Validate for graceful downgrade30 if response.sdk_http_response.headers.get("x-gemini-service-tier") == "standard":31 print("Warning: Priority limit exceeded, processed at Standard tier.")3233 print(response.text)3435 except Exception as e:36 # Standard error handling (e.g., DEADLINE_EXCEEDED)37 print(f"Error during API call: {e}")3839### JavaScript4041 import {GoogleGenAI} from '@google/genai';4243 const ai = new GoogleGenAI({});4445 async function main() {46 try {47 const result = await ai.models.generateContent({48 model: "gemini-3.8-flash",49 contents: "Triage this critical customer support ticket immediately.",50 config: {serviceTier: "priority"},51 });5253 // Validate for graceful downgrade54 if (result.sdkHttpResponse.headers.get("x-gemini-service-tier") === "standard") {55 console.log("Warning: Priority limit exceeded, processed at Standard tier.");56 }5758 console.log(result.text);5960 } catch (e) {61 console.log(`Error during API call: ${e}`);62 }63 }6465 await main();6667### Go6869 package main7071 import (72 "context"73 "fmt"74 "log"75 "google.golang.org/genai"76 )7778 func main() {79 ctx := context.Background()80 client, err := genai.NewClient(ctx, nil)81 if err != nil {82 log.Fatal(err)83 }84 defer client.Close()8586 resp, err := client.Models.GenerateContent(87 ctx,88 "gemini-3.8-flash",89 genai.Text("Triage this critical customer support ticket immediately."),90 &genai.GenerateContentConfig{91 ServiceTier: "priority",92 },93 )94 if err != nil {95 log.Fatalf("Error during API call: %v", err)96 }9798 // Validate for graceful downgrade99 if resp.SDKHTTPResponse.Header.Get("x-gemini-service-tier") == "standard" {100 fmt.Println("Warning: Priority limit exceeded, processed at Standard tier.")101 }102103 fmt.Println(resp.Text())104 }105106### REST107108 curl -X POST "https://generativelanguage.googleapis.com/v1beta/models/gemini-3.8-flash:generateContent?key=$GEMINI_API_KEY" \109 -H "Content-Type: application/json" \110 -d '{111 "contents": [{112 "parts":[{"text": "Analyze user sentiment in real time"}]113 }],114 "service_tier": "priority"115 }'116117## How Priority inference works118119Priority inference routes requests to high-criticality compute queues, offering120predictable, fast performance for user-facing applications. Its primary121mechanism is a graceful server-side downgrade to standard processing for traffic122that exceeds dynamic limits, ensuring application stability instead of failing123the request.124125| Feature | Priority | Standard | Flex | Batch |126|---|---|---|---|---|127| **Pricing** | 75-100% more than Standard | Full price | 50% discount | 50% discount |128| **Latency** | Seconds | Seconds to minutes | Minutes (1--15 min target) | Up to 24 hours |129| **Reliability** | High (Non-sheddable) | High / Medium-high | Best-effort (Sheddable) | High (for throughput) |130| **Interface** | Synchronous | Synchronous | Synchronous | Asynchronous |131132### Key benefits133134- **Low latency**: Designed for second response times for interactive, user-facing AI tools.135- **High reliability**: Traffic is treated with the highest criticality and is strictly non-sheddable.136- **Graceful degradation**: Traffic spikes exceeding dynamic limits are automatically downgraded to the Standard tier for processing instead of failing, preventing service outages.137- **Low friction** : Uses the same synchronous `generateContent` method as the standard and Flex tiers.138139### Use cases140141Priority processing is ideal for business-critical workflows where performance142and reliability are paramount.143144- **Interactive AI applications**: Customer service chatbots and copilots where users pay a premium and expect fast, consistent responses.145- **Real-time decision engines**: Systems requiring highly reliable, low-latency outcomes, such as live ticket triaging or fraud detection.146- **Premium customer features**: Developers who need to guarantee higher service level objectives (SLOs) for paying customers.147148### Rate limits149150Priority consumption holds its own rate limits even though consumption is151counted towards [overall interactive traffic rate limits](https://aistudio.google.com/rate-limit). The default rate limits152for Priority inference are **0.3x standard rate limit for Model / Tier**153154### Graceful downgrade logic155156If Priority limits are exceeded due to congestion, overflow requests are157**automatically and gracefully** downgraded to Standard processing instead of158failing with a 503 or 429 error. Downgraded requests are billed at the standard159rate, not the Priority premium rate.160161### Client responsibility162163- **Response monitoring** : Developers should monitor the `x-gemini-service-tier` header in the API response to detect if requests are being frequently downgraded to `standard`.164- **Retries** : Clients must implement retry logic/exponential backoff for standard errors, such as `DEADLINE_EXCEEDED`.165166## Pricing167168Priority inference is priced at 75-100% more than the [standard API](https://ai.google.dev/gemini-api/docs/pricing) and billed per token.169170## Supported models171172The following models support Priority inference:173174| Model | Priority inference |175|---|---|176| [Gemini 3.8 Flash](https://ai.google.dev/gemini-api/docs/models/gemini-3.8-flash) | ✔️ |177| [Gemini 3.7 Flash](https://ai.google.dev/gemini-api/docs/models/gemini-3.7-flash) | ✔️ |178| [Gemini 3.6 Flash](https://ai.google.dev/gemini-api/docs/models/gemini-3.6-flash) | ✔️ |179| [Gemini 3.5 Flash-Lite](https://ai.google.dev/gemini-api/docs/models/gemini-3.5-flash-lite) | ✔️ |180| [Gemini 3.5 Flash](https://ai.google.dev/gemini-api/docs/models/gemini-3.5-flash) | ✔️ |181| [Gemini 3.1 Flash-Lite](https://ai.google.dev/gemini-api/docs/models/gemini-3.1-flash-lite) | ✔️ |182| [Gemini 3.1 Pro Preview](https://ai.google.dev/gemini-api/docs/models/gemini-3.1-pro-preview) | ✔️ |183| [Gemini 3 Flash Preview](https://ai.google.dev/gemini-api/docs/models/gemini-3-flash-preview) | ✔️ |184| [Gemini 3 Pro Image Preview](https://ai.google.dev/gemini-api/docs/models/gemini-3-pro-image) | ✔️ |185| [Gemini 2.5 Pro](https://ai.google.dev/gemini-api/docs/models/gemini-2.5-pro) | ✔️ |186| [Gemini 2.5 Flash](https://ai.google.dev/gemini-api/docs/models/gemini-2.5-flash) | ✔️ |187| [Gemini 2.5 Flash Image](https://ai.google.dev/gemini-api/docs/models/gemini-2.5-flash-image) | ✔️ |188| [Gemini 2.5 Flash-Lite](https://ai.google.dev/gemini-api/docs/models/gemini-2.5-flash-lite) | ✔️ |189190## What's next191192Read about Gemini's other [inference and optimization](https://ai.google.dev/gemini-api/docs/optimization) options:193194- [Flex inference](https://ai.google.dev/gemini-api/docs/flex-inference) for 50% cost reduction.195- [Batch API](https://ai.google.dev/gemini-api/docs/batch-api) for asynchronous processing within 24 hours.196- [Context caching](https://ai.google.dev/gemini-api/docs/caching) for reduced input token costs.