Cost & Pricing
Controlling AI API cost and understanding pricing.
Articles in Cost & Pricing
- Provisioned Throughput vs Pay-As-You-Go: Gemini API Pricing for High-Volume Production — Pay-as-you-go charges per token with no ceiling on latency variance under load. Provisioned Throughput trades that flexibility for reserved, predictable capacity — at a cost structure worth understanding before you commit.
- Multimodal Input Costs on Gemini: How Images, Video, Audio, and PDFs Are Actually Tokenized and Priced — Every non-text input type has its own fixed, documented conversion into tokens before your model's per-token price ever applies. The exact rates for images, video, audio, and PDFs, with worked math for each.
- Monitoring and Capping Gemini API Spend: Budgets, Quotas, and Billing Alerts — Cost optimization only matters if you notice when it stops working. How to set up Google Cloud billing budgets, read usage from AI Studio, and decide whether a hard spend cap is worth the availability risk.
- Gemini API Pricing vs OpenAI vs Claude: A 2026 Cost Comparison — Cheapest tier, mid tier, and flagship tier compared across all three vendors' own pricing pages — with the batch and caching discounts each one offers layered on top.
- Google Search Grounding Costs: What Per-Request Grounding Actually Adds to Your Bill — Search grounding is priced per grounded prompt on Gemini 2.5 and per search request on Gemini 3.x — a distinction that changes how predictable the cost actually is. Real numbers and free allotments for both model lines.
- Gemini Developer API vs Vertex AI: Pricing and Billing Compared — Same underlying models, two different billing relationships — AI Studio pay-as-you-go with a real free tier versus Google Cloud project billing with enterprise features layered on top. What that actually changes about your bill.
- Gemini API Cost Optimization Playbook: Techniques That Actually Move the Bill — Seven independent levers for cutting a Gemini API bill — model tiering, caching, batching, thinking-level control, output trimming, selective tool use, and monitoring — and which ones to reach for first.
- How to Estimate Your Gemini API Bill Before You Build — A worked walkthrough for turning a feature idea into a real monthly cost estimate — counting tokens accurately, accounting for multimodal input, and modeling volume before you write production code.
- Gemini API Pricing 2026: Every Model's Cost Per Million Tokens Compared — One table for every current Gemini text model's per-million-token pricing, what the free tier actually includes, and the one promotional rate that expires at the end of this year.
- Gemini Thinking Level Guide: Controlling Reasoning Depth and Cost — Thinking tokens are billed as output tokens — so a "medium" default you never touch may be quietly overpaying for tasks that need low, or underdelivering on ones that need high. How the dial actually works.
- Gemini API Rate Limits Explained: RPM, TPM, RPD, and Usage Tiers — Rate limits are per-project, not per-key, and every model has its own RPM/TPM/RPD ceiling — so a second API key never buys you more headroom. How tier qualification actually works and how to handle a 429.
- Gemini Context Caching Guide: Implicit and Explicit Caching for Cost Control — Gemini 2.5+ models cache repeated prompt content automatically, for free — but understanding implicit vs explicit caching changes how you should structure prompts to actually hit the cache.
- Gemini Batch API Guide: Cutting LLM Costs 50% for Async Workloads — The Batch API trades a same-second response for a roughly 50% price cut, targeting completion within 24 hours. How to submit inline and file-based jobs, poll status, and know when it fits.
- Veo 3.1 Lite API Guide: Fast, Low-Cost Video Generation — Veo 3.1 Lite is Google's low-cost, low-latency video generation tier — up to 8x cheaper than Veo 3.1 Standard. Confirmed pricing, the async job/polling API pattern, and complete Python/JS code.
- Nano Banana 2 Lite Explained: Fast, Cheap Gemini Image Generation — Nano Banana 2 Lite trades peak fidelity for ultra-low latency and cost, making it the right choice for high-volume image generation. Full pricing breakdown and Python/JS code included.
- Gemini 2.5 Flash-Lite API Guide: Cheapest Multimodal Gemini Model — Gemini 2.5 Flash-Lite is the cheapest multimodal model in the 2.5 family. Here's how to build image+caption content moderation with the Gemini Interactions API, including pricing and full Python/JS code.
- Gemini 3.1 Flash-Lite API Guide: Frontier Performance at Minimum Cost — Gemini 3.1 Flash-Lite delivers frontier-class performance at the lowest cost in the active Gemini 3 line. Model ID, verified pricing, and multimodal Python/JS examples.
- Gemini 3.5 Flash-Lite API Guide: Fast, Cheap Gemini API Calls at Scale — Gemini 3.5 Flash-Lite is built for high-throughput, cost-sensitive workloads. Model ID, verified pricing, and Python/JS classification examples included.