Gemini API Call Price Analysis: 1 Million Characters Per Month in INR
Executive Summary
This report provides a comprehensive analysis of the Gemini API pricing structure, specifically addressing the cost implications for a monthly volume of 1 million characters, presented in Indian Rupees (INR). A fundamental aspect of Gemini API billing is its token-based model, not a character-based one. This necessitates a crucial conversion factor: approximately 1 million characters translate to 250,000 tokens per month.1
The cost for processing 1 million characters monthly varies significantly based on the selected Gemini model (e.g., Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.5 Flash-Lite) and the distribution between input and output tokens. The report establishes that output tokens consistently incur a substantially higher cost per unit compared to input tokens, making output length a primary driver of expenditure. While a precise, single cost figure is not feasible due to these variables, the provided scenario-based estimations offer a robust framework for financial planning. Additionally, the report clarifies the distinction between direct API billing and other Google AI consumer or developer subscriptions to prevent potential financial misinterpretations.
The billing mechanism for the Gemini API is centered on tokens, not characters. This is a critical distinction for accurate budgeting and understanding the underlying pricing structure. All cost analyses in this report are therefore framed around token usage, providing a precise representation of the billing model.
1. Understanding Gemini API Pricing Fundamentals
This section establishes the core concepts of the Gemini API's billing structure, essential for comprehending its cost implications.
1.1. Free Tier vs. Paid Tier: Benefits and Limitations
The Gemini API offers two primary pricing tiers: a "free of charge" tier and a "pay-as-you-go" tier. These tiers exhibit differences in pricing, rate limits, and the availability of specific models.4
The free tier is primarily intended for testing and development purposes, operating with lower rate limits. While certain models or usage within Google AI Studio may offer a 1 million token context window, this tier is generally not designed for the sustained, high-volume API usage implied by a query for 1 million characters per month, largely due to its inherent limitations.4
Conversely, the paid tier offers significant advantages, including substantially higher rate limits, access to additional features, and distinct data handling policies. For organizations, a particularly compelling aspect of the paid tier is its enhanced data privacy. When billing is enabled and the paid tier is utilized, prompts and responses are explicitly stated not to be used to improve Google products. In contrast, the free tier permits the use of data for product improvement.4 For entities managing sensitive, proprietary, or regulated information (such as customer data or confidential business strategies), the free tier's data usage policy presents a considerable compliance and security concern. The assurance provided by the paid tier regarding data privacy often becomes a non-negotiable requirement, making it the de facto choice for any production-grade application, regardless of the volume of usage. This transforms the decision to upgrade from a mere cost calculation to a strategic consideration of data governance.
To transition to the paid tier, users must enable Cloud Billing for their Google Cloud project, a process that can be conveniently initiated directly from Google AI Studio.4 It is important to note that usage within Google AI Studio itself remains free of charge, irrespective of whether billing is configured for API usage.4
1.2. Token-Based Billing: Input, Output, and Cached Tokens
The fundamental unit for Gemini API pricing is the token; all costs are calculated based on the number of tokens processed.2 Billing specifically accounts for four primary components:
Input token count: This refers to the tokens sent to the model as part of the prompt or request.4
Output token count: These are the tokens generated by the model as a response. This explicitly includes "thinking tokens" used by the model for internal computation, meaning users are not billed separately for these internal processes.1
Cached token count: This pertains to tokens stored for context caching, a feature designed to enable efficient reuse of information across multiple requests.4
Cached token storage duration: A recurring charge applied based on the volume of cached tokens and the length of time they are stored.4
Certain components do not incur charges. Requests made to the GetTokens API, which is used for pre-calculating token counts, are not billed and do not count against inference quota. Furthermore, if an API request fails with a 400 or 500 error, users are not charged for the tokens that would have been consumed, although the failed request will still count against their quota.1
The explicit billing for input, output, and cached tokens, combined with the provision of free token counting tools, indicates a design philosophy that empowers users to actively manage and predict their API costs. Google provides the countTokens API, which is free to use and has a high quota, for determining token counts before sending requests to the model. Additionally, the usageMetadata attribute is always returned in the response object without incurring any charge.1 The ability to set parameters like
thinking budget (for 2.5 models) and maxOutputTokens (for all Gemini models) is explicitly provided as a mechanism to control the number of tokens used and, consequently, manage costs.1 This comprehensive suite of tools and a transparent billing structure enable developers to move beyond merely reviewing past bills to proactively optimizing costs through careful prompt engineering, output constraint, and pre-computation of token usage. This approach fosters a culture of cost-aware AI development.
1.3. Character-to-Token Conversion for Gemini Models
The core conversion rate for Gemini models is approximately 1 token for every 4 characters.1 This ratio is fundamental for translating character-based usage estimates into the token-based billing model. For a general sense of scale, approximately 100 tokens are equivalent to 60-80 English words.1
Tokenization is the process by which text and other modalities are segmented into smaller units called tokens. These units can range from single characters to parts of words or entire words, depending on the model's vocabulary.1
While the query focuses on characters, implying text-based usage, Gemini is a multimodal model. All input and output, including images, video, and audio, are tokenized, but their conversion to tokens follows distinct rules:
Images: Converted based on their dimensions. Images with both dimensions less than or equal to 384 pixels are counted as 258 tokens each. Larger images are cropped and scaled into 768x768 pixel tiles, with each tile counting as 258 tokens.1
Video: Converted at a fixed rate of 263 tokens per second.1
Audio: Converted at a fixed rate of 32 tokens per second.1
The multimodal capabilities of Gemini introduce distinct cost structures that can lead to unforeseen expenses if not accounted for as an application evolves. For instance, if an application expands to incorporate image analysis, video processing, or audio transcription, the cost model shifts dramatically from character-based text processing. A budget based solely on text character conversion would significantly underestimate the actual cost for multimodal usage. This represents a critical variable that should be considered if the application's scope is likely to extend beyond pure text.
2. Key Conversion Rates for Cost Calculation
This section performs the essential conversions required to address the user's query in INR.
2.1. Converting 1 Million Characters to Tokens
Based on the established conversion rate of approximately 1 token for every 4 characters 1, a monthly volume of 1,000,000 characters translates directly into the following token count:
1,000,000 characters / 4 characters/token = 250,000 tokens per month.
This calculated volume of 250,000 tokens per month serves as the foundational basis for all subsequent cost estimations and scenario analyses within this report.
2.2. Current USD to INR Exchange Rate
For accurate conversion of USD-based Gemini API pricing into Indian Rupees, the most recent and consistent exchange rate will be applied. As per the available information, the rate indicates 1 United States Dollar (USD) = 87.66 Indian Rupees (INR).5
It is important to acknowledge that foreign exchange rates are dynamic and subject to daily fluctuations. The rate provided here is based on the available data and should be considered indicative for the purpose of this report. Actual billing will reflect the prevailing exchange rate at the time of the transaction.
3. Detailed Gemini API Model Pricing Analysis (Paid Tier)
This section provides specific cost details for various Gemini models under the paid tier, which is crucial for accurate budgeting.
3.1. Overview of Key Gemini Models and Pricing Tiers
Google offers a range of Gemini models, each optimized for different use cases and providing varying levels of capability and performance. The primary models relevant for general API calls and text generation include Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.5 Flash-Lite, and the earlier Gemini 2.0 Flash/Flash-Lite.5 All pricing for these models is consistently presented per 1 Million (1M) tokens in USD.5
A notable nuance for Gemini 2.5 Pro is its tiered pricing structure, which is contingent on the size of the input prompt. For prompts less than or equal to 200,000 tokens (<= 200k tokens), a lower input and output price applies. However, for prompts exceeding 200,000 tokens (> 200k tokens), a higher input and output price is incurred.5
While Gemini 2.5 Pro boasts a substantial context window of up to 2 million tokens 3, leveraging this capability with very long prompts (exceeding 200,000 tokens) results in a significant cost premium. For instance, the input price doubles from $1.25 to $2.50 per 1M tokens, and the output price increases by 50% from $10.00 to $15.00 per 1M tokens when prompts exceed this threshold.5 This tiered pricing creates a direct financial consideration for consistently pushing the upper limits of 2.5 Pro's long context window. Although the model is designed to handle large inputs, frequent use with prompts larger than 200,000 tokens substantially increases the per-token cost. For applications that require extensive context but are highly cost-sensitive, developers may need to consider strategies to keep prompts below the 200,000-token threshold or evaluate whether a more cost-effective model like Gemini 2.5 Flash, which does not have this tiered pricing, might suffice, even with a smaller context window. This highlights that raw capability does not always equate to cost-efficiency for all use cases.
3.2. Comparative Pricing Table: Input, Output, and Context Caching (USD & INR)
The following table serves as a central reference, offering a clear, side-by-side comparison of the per-million-token costs for various Gemini models across key billing components. All prices are presented in both USD and their INR equivalents, facilitating direct financial assessment. The conversion rate used is 1 USD = 87.66 INR.
Table 1: Gemini API Paid Tier Pricing per 1 Million Tokens (USD & INR)
This consolidated table directly addresses the need for comparative analysis by presenting all critical pricing information in a single, easy-to-read format with pre-calculated INR values. This enables rapid comparison of the cost-effectiveness of different models for specific input/output needs, making it a direct and actionable resource for decision-makers.
3.3. Other Billed Components: Grounding, Image Generation, Video Generation, Embeddings
Beyond core input and output tokens for text, the Gemini API supports various advanced features and modalities, each with its own pricing structure that can significantly influence overall costs.
Grounding with Google Search: This feature allows the model to retrieve and integrate real-time information from Google Search. It includes a free tier of 1,500 requests per day (RPD) for paid tier users (500 RPD for free tier, shared with Flash-Lite RPD). Beyond this free quota, requests are charged at $35 USD per 1,000 requests, equivalent to ₹3,068.10 per 1,000 requests.5
Image Generation (Imagen 4 Preview): Priced per image generated, not per token.
Imagen 4 Standard image price: $0.04 USD per image, equivalent to ₹3.51 per image.5
Imagen 4 Ultra image price: $0.06 USD per image, equivalent to ₹5.26 per image.5
Video Generation (Veo 3 Preview, Veo 2): Priced per second of video generated.
Veo 3 Preview (default): $0.75 USD per second, equivalent to ₹65.75 per second.5
Veo 2: $0.35 USD per second, equivalent to ₹30.68 per second.5
Gemini Embedding: Used for converting text into numerical vector representations for tasks like search, recommendation, or classification. The input price for embeddings is $0.15 USD per 1 Million tokens, equivalent to ₹13.15 per 1 Million tokens.5
Live API: Specific pricing applies to the Live API for models such as Gemini 2.5 Flash and Gemini 2.0 Flash, with distinct input and output rates for text, audio, and image/video streams.5
While the initial query focuses on character-based text, the availability of multimodal capabilities means that any expansion of the application's scope to include image or video generation, or even search grounding, introduces entirely new and potentially much higher cost centers that operate on different billing units. For example, generating a 1-minute video with Veo 3 Preview would cost $45 USD (60 seconds * $0.75/second), which is equivalent to generating 36 million input tokens using Gemini 2.5 Flash-Lite (at $0.10/M tokens). This highlights a significant potential for cost escalation if the application's scope expands without re-evaluating the billing model. It is therefore advisable to consider the potential for multimodal usage in future roadmaps and to factor in these distinct, non-token-based costs, as they can dramatically alter the overall expenditure beyond simple text processing.
4. Estimated Monthly Cost Scenarios for 1 Million Characters in INR
This section provides practical cost estimations based on the 250,000 tokens derived from 1 million characters, considering various input/output distributions and models.
4.1. Baseline Token Volume
As established in Section 2.1, a monthly volume of 1,000,000 characters is equivalent to 250,000 tokens per month. This total token count will be distributed between input (prompts, context) and output (model responses) for the scenario analysis.
4.2. Scenario-Based Cost Calculations
To provide a realistic range of potential costs, three common usage scenarios are analyzed based on varying ratios of input tokens to output tokens. Calculations are performed for three representative Gemini models: Gemini 2.5 Pro (assuming prompts consistently stay below 200,000 tokens to reflect the lower pricing tier), Gemini 2.5 Flash, and Gemini 2.5 Flash-Lite.
Scenario 1: Input-Heavy Usage (70% Input / 30% Output)
This scenario represents applications where prompts are detailed, or conversations involve significant context, but responses are relatively concise (e.g., summarization, data extraction from large documents).
Input tokens: 250,000 * 0.70 = 175,000 tokens
Output tokens: 250,000 * 0.30 = 75,000 tokens
Scenario 2: Balanced Usage (50% Input / 50% Output)
This scenario represents a typical conversational AI or content generation application where input and output lengths are roughly balanced (e.g., chatbots, interactive content creation).
Input tokens: 250,000 * 0.50 = 125,000 tokens
Output tokens: 250,000 * 0.50 = 125,000 tokens
Scenario 3: Output-Heavy Usage (30% Input / 70% Output)
This scenario represents applications where prompts are brief, but the generated responses are extensive (e.g., long-form content generation, detailed explanations from short queries).
Input tokens: 250,000 * 0.30 = 75,000 tokens
Output tokens: 250,000 * 0.70 = 175,000 tokens
A consistent trend across all Gemini models is that the cost per output token is significantly higher than the cost per input token. For example, Gemini 2.5 Flash-Lite's output token cost is $0.40/M tokens, while its input token cost is $0.10/M tokens.5 Similarly, for Gemini 2.5 Pro (prompts <= 200k), the output cost is $10.00/M tokens compared to an input cost of $1.25/M tokens.5 This disparity means that even a modest increase in the average length of model responses (output tokens) will have a disproportionately larger impact on the total monthly bill compared to an equivalent increase in input tokens. Therefore, strategies focused on constraining or optimizing output length, such as utilizing the
maxOutputTokens parameter 1, are far more effective for cost reduction than solely focusing on prompt brevity.
Example Calculation (Gemini 2.5 Flash-Lite, Input-Heavy Scenario):
For the Input-Heavy scenario (70% Input / 30% Output) with Gemini 2.5 Flash-Lite:
Total tokens: 250,000
Input tokens: 175,000
Output tokens: 75,000
From Table 1:
Gemini 2.5 Flash-Lite Input Price: ₹8.77 per 1M tokens
Gemini 2.5 Flash-Lite Output Price: ₹35.06 per 1M tokens
Calculations:
Input cost: (175,000 / 1,000,000) * ₹8.77 = 0.175 * ₹8.77 = ₹1.53
Output cost: (75,000 / 1,000,000) * ₹35.06 = 0.075 * ₹35.06 = ₹2.63
Total Estimated Monthly Cost: ₹1.53 + ₹2.63 = ₹4.16
4.3. Estimated Monthly Cost Table (INR)
The following table consolidates the calculated monthly costs for each model under the different input/output distribution scenarios, providing a clear and comprehensive view of potential expenditures in INR.
Table 2: Estimated Monthly Gemini API Costs for 1 Million Characters (250,000 Tokens) in INR
This table directly addresses the user's query by providing concrete, actionable financial figures. It allows for rapid estimation of monthly expenditure based on anticipated usage patterns and chosen models, which is critical for budgeting and financial planning. By presenting costs across different scenarios and models, the table clearly demonstrates the significant variability in potential expenses, preventing the assumption of a single, fixed cost and promoting transparency in cost management.
5. Additional Cost Considerations and Nuances
This section expands on other factors that can influence overall Gemini API costs beyond basic input/output tokens, ensuring a comprehensive understanding of the billing landscape.
5.1. Context Caching and Storage Costs
Context caching is a feature designed to optimize performance and potentially reduce costs by storing large prompts or conversational history. By referencing cached tokens, the need to send the full context with every request is reduced, thereby lowering repetitive input token charges.4
Context caching involves two distinct billing components:
Context caching price: A one-time charge per 1 Million tokens for the act of caching the content (e.g., $0.31 USD per 1M tokens for Gemini 2.5 Pro prompts <= 200k).5
Context caching storage price: A recurring charge based on the volume of cached tokens and the duration they are stored, typically expressed per 1 Million tokens per hour (e.g., $4.50 USD per 1,000,000 tokens per hour for Gemini 2.5 Pro).5
Context caching is generally not available in the free tier.5 While context caching offers potential savings on input tokens by reducing redundancy, its own per-token caching and recurring storage costs introduce a new layer of financial consideration. For applications with short-lived contexts, infrequent context reuse, or very large cached contexts, the cumulative storage costs might quickly outweigh the savings from reduced input tokens. This means that a careful cost-benefit analysis, based on specific application data retention and access patterns, is necessary to determine if caching truly results in a net cost reduction. It represents a strategic decision that must balance performance benefits with sustained operational costs.
5.2. Distinction: Gemini API Billing vs. Consumer/Developer Subscriptions
It is imperative to differentiate between the direct, pay-as-you-go billing for Gemini API calls, which is the focus of this report, and other Google AI-related subscriptions or bundled offerings. Conflating these can lead to significant financial miscalculations.
Gemini API Pay-as-you-go Billing: This is the primary method for charging for direct API usage. Costs are incurred based on the actual consumption of input, output, and cached tokens, and are managed through Google Cloud Billing.4
Consumer Subscriptions (e.g., Google AI Pro, Google AI Ultra): These are typically monthly subscription plans (e.g., $19.99 USD/month for Google AI Pro, $249.99 USD/month for Google AI Ultra). They are primarily designed for individual consumers, offering enhanced features for the Gemini app, NotebookLM, increased storage (e.g., 2 TB), and other Google services. While some plans might include limited access or credits (e.g., 1,000 monthly AI credits for Flow and Whisk in Google AI Pro 10), they are not the direct billing mechanism for Gemini API calls themselves.10
Developer Program Subscriptions (e.g., Premium, Enterprise): These are yearly or monthly subscriptions (e.g., $299/year for Premium) aimed at developers. They provide access to tools like Gemini Code Assist, Gemini CLI, Firebase Studio workspaces, Google Cloud credits (e.g., $550 annual credit for Premium, $150 monthly credit for Enterprise), and other developer benefits. While these subscriptions offer credits that can be applied towards GenAI and Cloud services, the underlying API calls still fall under the standard pay-as-you-go Cloud Billing structure.12
The proliferation of "Gemini" branded subscriptions can easily lead to a misunderstanding that purchasing such a plan covers direct API usage, when in fact, API calls are separately billed. A user might mistakenly purchase a "Google AI Pro" subscription assuming it covers their production API usage, only to receive a separate, significant bill from Google Cloud for their actual API calls. This distinction is paramount for accurate budgeting and avoiding unexpected costs. These subscriptions are designed to provide benefits and credits, not direct API usage coverage.
5.3. Billing Setup and Usage Monitoring
The Gemini API's paid tier operates entirely within the Google Cloud Billing system. This integration means that all API usage charges will appear on the user's Google Cloud bill.4
After enabling Cloud Billing, users gain access to comprehensive monitoring tools within the Google Cloud console. These tools allow for detailed tracking of API usage, understanding cost breakdowns, making payments, and accessing Cloud Billing support. The service name for the API within the console is generativelanguage.googleapis.com.4
The deep integration of Gemini API billing with the robust Google Cloud Billing system provides powerful capabilities for proactive financial management. This is essential for controlling costs in dynamic AI applications. For a technical decision-maker, this means having granular, near real-time visibility into consumption patterns, enabling the identification of usage spikes, pinpointing costly operations, analyzing trends, and forecasting future expenditures with greater accuracy. This transforms cost management from a reactive process of reviewing past bills into a proactive, data-driven strategy for optimizing resource allocation and ensuring budget adherence. Any existing Google Cloud credits associated with the user's account can be applied towards Gemini API usage, potentially reducing out-of-pocket expenses.4
6. Recommendations for Cost Optimization
This section provides actionable advice for minimizing Gemini API expenditures while maximizing utility, drawing on the analysis of the pricing structure.
6.1. Strategic Model Selection
The most impactful cost optimization strategy involves selecting the Gemini model that precisely matches the functional requirements of the task, rather than defaulting to the most powerful model.
Gemini 2.5 Flash-Lite: This is the most cost-effective option for simpler text generation tasks, basic summarization, or applications where speed and minimal cost are the highest priorities.5
Gemini 2.5 Flash: Offers a strong balance of capability and cost-efficiency, making it a suitable general-purpose model for a wide range of applications that require good performance without the premium cost of the Pro model.5
Gemini 2.5 Pro: Should be reserved for highly complex tasks that genuinely require its advanced reasoning capabilities, larger context window, or multimodal understanding. For cost efficiency, efforts should be made to keep individual prompt sizes below the 200,000 token threshold to avoid the higher pricing tier for input and output.5
Using a more powerful and expensive model than what is functionally necessary for a given task is a direct and often overlooked path to unnecessary expenditure. The significant price differences between models (e.g., Gemini 2.5 Pro output at $10.00/M tokens versus Gemini 2.5 Flash-Lite output at $0.40/M tokens) 5 mean that if a task can be adequately performed by Gemini 2.5 Flash-Lite, choosing Gemini 2.5 Pro would result in a 25-fold increase in output token cost for the same functional outcome. This represents substantial over-provisioning of computational resources and a direct waste of budget. Understanding this trade-off is crucial for optimizing resource allocation.
6.2. Optimizing Token Usage
Effective Prompt Engineering: Craft concise, clear, and efficient prompts to minimize the number of input tokens required to convey instructions or context to the model without compromising the quality of the desired output.
Controlling Output Length: Given that output tokens are significantly more expensive than input tokens, explicitly limiting the length of model responses is a critical optimization. This can be achieved by utilizing the maxOutputTokens parameter in the model's configuration.1 Unconstrained model responses, even if seemingly useful, can lead to significant cost inflation due to the high price of output tokens. Without explicit instructions, generative AI models can be verbose, providing more detail or conversational filler than strictly necessary. This inherent verbosity, if unchecked, directly translates into higher token consumption and thus higher costs. The
maxOutputTokens parameter provides a direct mechanism to mitigate this, forcing the model to be more concise and focused, thereby directly impacting the bill.Thinking Budget (for 2.5 models): For Gemini 2.5 models, setting a "thinking budget" can further help control costs by limiting the internal processing tokens the model uses, which are included in the output price.1
Leveraging countTokens API: Proactively use the free countTokens API 1 to estimate the token count of prompts and potential responses
before making actual API calls. This allows for iterative refinement of prompts and output constraints to optimize token usage pre-deployment.
6.3. Monitoring and Managing API Spending
Utilize Google Cloud Console: Regularly monitor Gemini API usage and spending patterns within the Google Cloud console. The console provides detailed reports and dashboards that can help identify trends, anomalies, and potential areas of overspending.4
Set Budgets and Alerts: Configure Cloud Billing budgets and alerts to receive automated notifications when spending approaches predefined thresholds. This proactive alerting mechanism helps prevent unexpected high bills.
Review Quotas: Understand and manage API quotas. While quotas are distinct from billing, exceeding them can lead to service interruptions. Aligning quota requests with anticipated usage and budget is crucial for smooth operation and cost control.4
7. Conclusion
This report has provided a comprehensive analysis of the Gemini API pricing structure, specifically addressing the cost of generating 1 million characters per month in Indian Rupees. It has been established that Gemini API billing is fundamentally token-based, not character-based, with approximately 1 million characters equating to 250,000 tokens.
The cost of Gemini API calls is highly variable, primarily dependent on the specific Gemini model chosen (e.g., 2.5 Pro, 2.5 Flash, 2.5 Flash-Lite), the distribution of input versus output tokens within API calls, and the utilization of additional features such as context caching or multimodal inputs (image/video generation, search grounding). Output tokens consistently represent a significantly higher cost per unit than input tokens, making output length optimization a critical factor.
While a precise single figure for 1 million characters is elusive due to these variable factors, the provided scenario-based cost estimations offer a robust framework for accurate budgeting and financial planning. Furthermore, the crucial distinction between direct API billing and other Google AI consumer/developer subscriptions has been clarified to prevent potential financial misinterpretations.
Given the inherent variability in AI API pricing, usage patterns, and the continuous evolution of models and features, effective cost management is not a one-time calculation but an ongoing, iterative process. The report has detailed multiple variables influencing cost, including model choice, input/output ratio, multimodal usage, caching, tiered pricing for Pro models, and fluctuating exchange rates.5 Additionally, preview models may change before becoming stable and have more restrictive rate limits 5, indicating that pricing and capabilities are not static. Relying on a static, initial cost estimate for a dynamic AI application is insufficient. Successful integration and sustained cost-effectiveness require continuous monitoring of actual usage against budgeted amounts, adaptation of optimization strategies as application needs evolve, and staying informed about changes in Google's pricing and model offerings. This elevates cost management from a simple accounting task to a strategic, ongoing operational imperative for any organization leveraging generative AI at scale.
Works cited
Count tokens for Gemini models | Firebase AI Logic - Google, accessed August 6
, 2025, https://firebase.google.com/docs/ai-logic/count-tokens
Understand and count tokens | Gemini API | Google AI for Developers, accessed August 6, 2025,
How many characters is allowed in each prompt - Google AI Developers Forum, accessed August 6, 2025,
https://discuss.ai.google.dev/t/how-many-characters-is-allowed-in-each-prompt/79901
Billing | Gemini API | Google AI for Developers, accessed August 6, 2025,
Gemini Developer API Pricing | Gemini API | Google AI for Developers, accessed
August 6, 2025,
Understand pricing | Firebase AI Logic - Google, accessed August 6, 2025,
Gemini API – APIs & Services - Google Cloud Console, accessed August 6, 2025,
https://console.cloud.google.com/apis/library/generativelanguage.googleapis.com
Currency watch: Rupee rebounds 15 paise to 87.65 against US dollar after all-time
low; RBI action, crude dip offer support, accessed August 6, 2025,
Long context | Gemini API | Google AI for Developers, accessed August 6, 2025,
Google AI Plans and Features, accessed August 6, 2025,
Google AI Pro & Ultra — get access to Gemini 2.5 Pro & more, accessed August 6,
Google Developer Program Plans & Pricing, accessed August 6, 2025,