API model string
gemini-2.5-flashContext window
1M tokens
Max output 65K tokens
- Knowledge cutoff
- not published
- Released
- 2025
- Modalities
- text, image, audio, video in; text out
Last verified July 10, 2026
Rate limits by tier
Google assigns rate limits per project and no longer publishes a universal RPM/TPM table. Your active quotas are visible in the AI Studio console. All tiers share the same model endpoint (gemini-2.5-flash); billing status determines which tier applies.
| Tier | Requirements | RPM | TPM | RPD | Concurrent | Notes |
|---|---|---|---|---|---|---|
| Free (AI Studio) | Google account; no credit card required | not published (project-specific) | not published | ~1,500 (approximate; not guaranteed — verify in AI Studio console) | not published | Free tier data may be used to improve Google products. Rate limits shown per project in AI Studio, not in public docs. |
| PAYG (Pay-as-you-go) | Google Cloud billing account linked to project | not published (project-specific, shown in AI Studio) | not published | not published | not published | $0.30 in / $2.50 out per MTok (raised July 2, 2026). No long-context surcharge — flat rate across all context lengths. Paid-tier data is not used for training. |
| Batch API | Google Cloud billing account | async (results within 24h) | not published | not published | not published | 50% discount: $0.15 in / $1.25 out per MTok. For non-latency-sensitive workloads: bulk classification, document processing, overnight jobs. |
| Vertex AI | GCP project with Vertex AI API enabled | not published (project-specific via Vertex quotas) | not published | not published | not published | Separate Vertex pricing applies; a 10% non-global endpoint premium took effect July 1, 2026. Use for VPC, data residency, and enterprise SLA requirements. |
| Enterprise | Google Cloud Sales agreement | negotiated | negotiated | negotiated | negotiated | Dedicated capacity, committed-use discounts, custom SLAs. Contact cloud.google.com/contact. |
Swipe the table sideways to see every limit column.
- 1.Google no longer publishes a stable RPM/TPM table for Gemini models. Active limits are project-specific and shown in the AI Studio console — the docs note that published limits are not guaranteed.
- 2.Thinking tokens are billed at the output rate ($2.50/MTok). Disabling thinking mode prevents unexpected output-volume spikes.
- 3.No long-context surcharge: Gemini 2.5 Flash uses flat pricing at all context lengths up to 1M tokens, unlike Gemini 2.5 Pro which doubles input cost above 200K tokens.
Limits verified against the Google docs, July 10, 2026.
Token pricing
What you pay per million tokens (USD). Input and output are billed separately.
Input
$0.30
per 1M tokens
Output
$2.50
per 1M tokens
Cached input
$0.03
per 1M cached
Batch discount
−50%
on batch jobs
- Prices raised July 2, 2026. Previous pricing not published by Google.
- Context caching: cache reads billed at $0.03/MTok (~10% of standard input). Cache write cost varies — check AI Studio for current cache storage pricing.
- Thinking tokens bill at the output rate ($2.50/MTok) and can multiply output volume 2–5×. Disable thinking for tasks that do not require step-by-step reasoning.
- Batch API: 50% off all input/output — $0.15/$1.25 per MTok. Results delivered within 24h.
- Free tier: data may be used by Google to improve models. PAYG tier: data is not used for training.
- Vertex AI pricing differs from Generative Language API pricing; a 10% non-global endpoint premium applies on Vertex as of July 1, 2026.
Side-project chatbot
$4.00
per month
Assumptions
5M input tokens / 1M output tokens per month
5M × $0.30/1M + 1M × $2.50/1M = $1.50 + $2.50
Document processing (70% cached reads)
$36.66
per month
Assumptions
60M input tokens / 12M output tokens per month; 42M served from cache
(18M × $0.30 + 42M × $0.03)/1M + 12M × $2.50/1M = $5.40 + $1.26 + $30.00
Enterprise scale
$320
per month
Assumptions
400M input tokens / 80M output tokens per month
400M × $0.30/1M + 80M × $2.50/1M = $120 + $200
Run your own numbers
Drag your real monthly token volumes and watch the bill update live — priced against rival models at the same usage.
Estimated gemini-2.5-flash spend
$22/mo
Input: $9.00
Output: $13
30M in × $0.300 + 5M out × $2.50 = $22
Same volume, priced across models
- Gemma 3Cheapest$1.60
- gemini-2.5-flashThis model$22
- Gemini 2.5 Pro$88
- Claude 4 Sonnet$110
Rivals priced at their published input/output rates for the same monthly volumes. Prompt caching is model-specific, so it is applied to gemini-2.5-flash only. Estimates for comparison; real bills vary with request shape and long-context surcharges.
gemini-2.5-flash vs the alternatives
Gemini 2.5 Flash is the best-value closed-model option in Google's lineup — cheaper than Gemini 2.5 Pro on both dimensions, with a free tier and no long-context surcharge.
| Aspect | gemini-2.5-flash | Gemini 2.5 Pro | Claude 4 Sonnet | Gemma 3 |
|---|---|---|---|---|
| Context window | 1M tokens | 1M+ tokens (Pro) | 1M tokens (Claude 4 Sonnet) | 131K tokens (Gemma 3) |
| Max output | 65K tokens | not published (Pro) | 128K tokens (Sonnet) | not published (Gemma 3) |
| Input $/MTok | $0.30 | $1.25 (Pro) | $2.00 (Sonnet intro) | $0.040 (Gemma 3 4B hosted) |
| Output $/MTok | $2.50 | $10.00 (Pro) | $10.00 (Sonnet intro) | $0.080 (Gemma 3 4B hosted) |
| Long-context surcharge | None — flat pricing | Yes, above 200K (Pro, whole prompt) | None (Sonnet) | None (Gemma 3) |
| Free tier | Yes (~1,500 RPD approximate) | None (Pro — paid only since Apr 1, 2026) | None (Sonnet) | Yes (Gemma 3 AI Studio) |
| Batch discount | 50% | 50% (Pro) | 50% (Sonnet) | 50% (Gemma 3 hosted) |
| Multimodal input | Text, image, audio, video | Text, image, audio, video (Pro) | Text, image (Sonnet) | Text, image (Gemma 3) |
Swipe the table sideways to see every model.
Hitting a 429? The playbook
The exact errors you'll see
429 Too Many RequestsRESOURCE_EXHAUSTEDQuota exceeded for quota metric 'generate_requests_per_minute'You have exceeded your current quota, please check your plan and billing detailsWhy it happens & how to fix it
Free-tier daily request limit (~1,500 RPD) reached
Spread requests across the 24h window, or add a Google Cloud billing account to upgrade to PAYG with higher project quotas.
Project RPM quota exceeded
Implement a request queue with controlled concurrency. Use exponential backoff with jitter. View your current RPM quota in AI Studio and request an increase in Google Cloud Console → IAM & Admin → Quotas.
No billing account while attempting high-volume use
Add a billing account at console.cloud.google.com. The PAYG rate is $0.30/$2.50 with no minimum spend. Billing eliminates the restrictive free-tier RPD cap.
Thinking tokens causing unexpected quota consumption
Disable thinking mode for tasks that do not require step-by-step reasoning. Thinking can multiply output token volume 2–5×, exhausting output quotas faster.
Concurrent request spike exceeding project limits
Add request queuing at the application layer. Process a fixed number of concurrent requests (e.g., 5–10) with a semaphore or queue. Respect the Retry-After header returned by the API.
Retry strategy
Google AI docs explicitly recommend exponential backoff. Start with a 1-second delay, double on each retry, cap at 64 seconds. Add ±20% random jitter to avoid thundering-herd. Use the Retry-After header when present. Google's official client libraries (google-generativeai for Python, @google/generative-ai for Node.js) include built-in retry logic.
1// retry.ts — Gemini 2.5 Flash with exponential backoff2const API_KEY = process.env.GOOGLE_AI_API_KEY!;3const BASE_URL = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent";45async function generateWithRetry(6 prompt: string,7 maxRetries = 58): Promise<string> {9 let delay = 1000; // ms10 for (let attempt = 0; attempt <= maxRetries; attempt++) {11 const res = await fetch(`${BASE_URL}?key=${API_KEY}`, {12 method: "POST",13 headers: { "Content-Type": "application/json" },14 body: JSON.stringify({15 contents: [{ parts: [{ text: prompt }] }]16 })17 });1819 if (res.ok) {20 const data = await res.json();21 return data.candidates[0].content.parts[0].text;22 }2324 if (res.status === 429 && attempt < maxRetries) {25 const retryAfter = res.headers.get("Retry-After");26 const wait = retryAfter27 ? parseInt(retryAfter, 10) * 100028 : delay * (1 + Math.random() * 0.4 - 0.2); // ±20% jitter29 console.warn(`429 — retrying in ${Math.round(wait / 1000)}s (attempt ${attempt + 1})`);30 await new Promise((r) => setTimeout(r, wait));31 delay = Math.min(delay * 2, 64000);32 continue;33 }3435 throw new Error(`Gemini API error ${res.status}: ${await res.text()}`);36 }37 throw new Error("Max retries exceeded");38}How to raise your limits
The ladder from the starter tier to enterprise — what each rung takes, and what it unlocks.
Free (AI Studio)
ImmediateSign in at aistudio.google.com with a Google account. No credit card needed.
Unlocks: ~1,500 RPD (approximate, not guaranteed). Good for prototyping and low-volume evaluation.
PAYG
Minutes after billing is enabledGoogle Cloud Console → create or link a billing account → associate it with your AI Studio project.
Unlocks: Pay-per-token at $0.30/$2.50 per MTok. Higher project quotas. Data not used for Google model training.
Higher RPM / TPM quotas
1–5 business daysGoogle Cloud Console → IAM & Admin → Quotas → Generative Language API → find your quota → click Request increase. Provide expected volume and use case.
Unlocks: Custom per-project RPM and TPM limits beyond default PAYG caps.
Vertex AI
Same dayEnable the Vertex AI API on your GCP project. Use the Vertex AI SDK (Python: google-cloud-aiplatform; Node: @google-cloud/aiplatform). See cloud.google.com/vertex-ai/pricing.
Unlocks: Enterprise SLA, VPC Service Controls, data residency, committed-use pricing. Note: 10% non-global endpoint premium as of July 1, 2026.
Enterprise
NegotiatedContact Google Cloud Sales at cloud.google.com/contact.
Unlocks: Dedicated capacity, custom rate limits, higher SLAs, and committed-use discounts.
Cut your token spend
Context caching
−90% input cost on repeated prefixesCache system prompts, document prefixes, or static context once and reuse. Cache reads cost $0.03/MTok — 10% of the standard $0.30 input rate. Critical for long-context pipelines that reuse the same prefix across many requests.
Stay on the free no-surcharge path
Zero long-context cost penaltyGemini 2.5 Flash charges flat pricing across all context lengths up to 1M tokens. Unlike Gemini 2.5 Pro (which doubles input cost above 200K), Flash never imposes a surcharge — use the full context window without budget risk.
Batch API for async workloads
−50% cost on all tokensSubmit bulk jobs via the Batch API. Results return within 24h. Rates: $0.15/$1.25 per MTok. Ideal for document classification, bulk extraction, and overnight data processing pipelines.
Disable thinking for simple tasks
2–5× reduction in output token volumeThinking tokens are billed at the output rate ($2.50/MTok) and can multiply the token count of a response significantly. Use thinking only when step-by-step reasoning meaningfully improves results.
Use Flash instead of Pro for appropriate workloads
75–80% lower cost vs Gemini 2.5 ProFlash at $0.30/$2.50 vs Pro at $1.25/$10.00 on the same 1M context window. Use Pro only for tasks that genuinely require its superior reasoning; route everything else to Flash.
Free tier for prototyping only
Avoids quota unpredictability in productionThe free-tier RPD limit (~1,500 RPD approximate) is not published or guaranteed. Switch to PAYG before launching production workloads to avoid hit-or-miss quota behavior.
Monitor token counts per request
Prevents unexpected truncation at 1M token limitFlash's 1M context window is large but not infinite. Log input token counts via the API usage metadata and alert when approaching 800K tokens to avoid truncation or request rejection.
Frequently asked questions
How do I increase Gemini 2.5 Flash rate limits?
Open Google Cloud Console → IAM & Admin → Quotas → search for 'Generative Language API' → find your quota metric → click Request increase. Provide your expected volume and use case. Google typically responds in 1–5 business days. You must have a billing account linked to your project before you can request an increase.
Is Gemini 2.5 Flash free to use?
Yes, there is a free tier via AI Studio with approximately 1,500 requests per day (not a guaranteed limit — verify in your console). The free tier is sufficient for prototyping. For production, upgrade to PAYG at $0.30/$2.50 per MTok; the minimum spend is $0 and there is no subscription fee.
What does RESOURCE_EXHAUSTED mean on Gemini 2.5 Flash?
RESOURCE_EXHAUSTED is Google's error code for a 429 rate-limit hit. It means your project has exceeded its RPM or RPD quota. Implement exponential backoff (start 1s, double each retry, cap 64s, add ±20% jitter), check your quotas in AI Studio, and request a quota increase in Google Cloud Console if you need higher sustained throughput.
Gemini 2.5 Flash vs Gemini 2.5 Pro — which should I use?
Use Flash for most production workloads. Flash is $0.30/$2.50 per MTok vs Pro's $1.25/$10.00 — roughly 75–80% cheaper — and Flash has no long-context surcharge while Pro doubles input cost for prompts over 200K tokens. Choose Pro only when you need its stronger reasoning for complex tasks like multi-step analysis, advanced coding, or difficult Q&A.
Does Gemini 2.5 Flash have a long-context price surcharge?
No. Gemini 2.5 Flash uses flat pricing across all context lengths up to 1M tokens. Gemini 2.5 Pro, by contrast, charges $2.50 input (vs $1.25) and $15.00 output (vs $10.00) when any part of the prompt exceeds 200K tokens — and the surcharge applies to the entire prompt, not just the tokens above the threshold. Flash eliminates this risk entirely.
How are thinking tokens priced on Gemini 2.5 Flash?
Thinking tokens are billed at the standard output rate of $2.50 per MTok. When thinking mode is active, the model generates an internal chain-of-thought before the final response, which can multiply total output token volume 2–5×. Disable thinking for tasks where you do not need step-by-step reasoning to keep costs predictable.
What is the Gemini 2.5 Flash context cache price?
Context cache reads cost approximately $0.03 per MTok — about 10% of the standard $0.30 input price. If your pipeline reuses the same long system prompt or document prefix across many requests, caching can reduce per-request input costs by up to 90% on the cached portion.
Can RapidDev help with Gemini 2.5 Flash quota and cost engineering?
Yes. If you're hitting quota walls or want to optimise your Gemini spend, RapidDev specialises in API gateway design, quota management, and token-cost reduction. Book a free scoping call at rapidevelopers.com/contact.
We build AI apps that don't hit rate limits
- Retry, backoff & caching built in
- Multi-provider fallback routing
- Fixed price, you own the code
30-min call. No commitment.