Gemini API Review 2026

What the Gemini API actually is

The Gemini API is the developer door to Google DeepMind models, run by Google from Mountain View and fronted by Google AI Studio, the browser console where keys, projects, usage and billing live. It launched in December 2023 and the documentation moved from ai.google.dev to aistudio.google.com/docs during 2026. It’s distinct from Vertex AI, the enterprise path on Google Cloud, although the same project can use both.

That split is the whole credits story. AI Studio bills through a prepaid or postpaid balance attached to a Cloud Billing account, while Vertex AI bills as ordinary Google Cloud usage. Credits that attach to a billing account, such as the Welcome credit or startup program grants, flow to Vertex AI but, since March 2026, not to AI Studio balances. Same model, two doors, and only one of them takes credits.

What free gets you and what it costs

There’s no trial credit. Instead, every active project starts on the free tier, where Gemini 3.8 Flash, 3.6 Flash, 3.5 Flash-Lite, 3.1 Flash-Lite, 3 Flash Preview, Gemini Embedding 2 and gemini-embedding-001 are free of charge for both input and output. Gemini 3.1 Pro Preview, Gemini 4 Argon and Gemini Omni Flash aren’t available on the free tier at all.

The price of free is data. Free tier content can be used to improve Google products and human reviewers may read it. In the EEA, UK and Switzerland the paid data terms cover free usage too. The only outside credits that reach Gemini usage are the $10 or $100 a month that Google AI Pro and Ultra subscribers have received since the January 27, 2026 announcement, which can be applied to the Gemini API, plus program credits spent via Vertex AI.

How rate limits actually bite

Rate limits are enforced per project along three dimensions: requests per minute, input tokens per minute and requests per day, with some models adding images per minute or tokens per day. Daily quotas reset at midnight Pacific. Exceed any one and you get a 429. The public rate limits docs no longer carry a per-model table for the free tier and point to the rate limit page inside AI Studio, which requires sign-in.

That opacity is the main weakness. Forum threads record a sharp cut in December 2025, when a Google representative said the documentation had been updated. Since September 18, 2026 access to Gemini 2.5 models is limited to projects that already used them, and in November 2026 Google deprecated 2.5 Pro and 2.5 Flash with automatic routing to 3.8 Flash and 3.5 Flash-Lite from December 1. Plan on the free tier changing under you.

What it costs once you pay

Standard paid prices per million tokens: Gemini 3.1 Flash-Lite $0.25 input and $1.50 output; 3.5 Flash-Lite $0.30 and $2.50; 3.8 Flash and 3.6 Flash $0.75 and $3.75 through December 31, 2026, then $1.50 and $7.50; 3.1 Pro Preview $2.00 and $12.00 up to 200,000 tokens of context; Gemini 4 Argon $2.00 and $10.00 through November 30, 2026, then $4.00 and $20.00. Batch is half price, Flex is cheaper than Standard and Priority costs more.

Billing is set at the billing account level. Tier 1 begins when you link a billing account and carries a $250 monthly cap and a $10 limit per rolling 10 minutes. Tier 2 requires $100 paid plus 3 days and caps at $2,000. Tier 3 requires $1,000 plus 30 days and caps between $20,000 and more than $100,000. Prepay, rolled out March 23, 2026, has a $5 minimum and $5,000 maximum balance, and keys stop with HTTP 402 at zero.

Credits for startups and researchers

Google has no startup program specific to the Gemini API. The route is the Google for Startups Cloud Program, whose Scale tier covers 100% of Google Cloud usage up to $100,000 in year one, or $250,000 for AI-first startups, and whose credits cover Gemini and Gemma. Those credits are spent on Vertex AI SKUs, so a startup that built on an AI Studio key has to switch endpoints to use them. Do that before you apply, not after.

The Start tier for pre-funded companies grants $2,000 and lists free tier AI Studio access as a benefit, which is the same access everyone has. Google Cloud research credits of $5,000 and student Google Skills credits exist on the education side but aren’t Gemini-specific. For an individual, the practical subsidy is a Google AI Pro subscription with its $10 monthly Cloud credit.

Models and features that change the bill

The model lineup turned over fast in 2026. Gemini 3.1 Pro Preview arrived February 19, 3.1 Flash-Lite went GA May 7, 3.5 Flash on May 19, 3.6 Flash and 3.5 Flash-Lite on July 21, Omni Flash on August 27, 3.8 Flash on September 2 and Gemini 4 Argon on October 13. Pricing moved with each release, and two introductory periods end within weeks of each other at the turn of the year.

Features that change the bill: context caching at a tenth of input price plus storage, Grounding with Google Search billed only on requests that return a web URL and capped at 1,500 a day by default, Batch at 50% off, and the Flex and Priority tiers introduced April 1, 2026. The temperature, top_p and top_k parameters were deprecated on July 21, 2026, which can break older sampling-based cost tricks.

The ways it can stop on you

Support on the free tier is the developer forum. Paid tiers get billing support through the Cloud console. The risks are operational rather than financial: a prepay balance at $0 stops every key on the billing account, a tier cap pauses all linked projects, and the 2.5 deprecation shows that model routing can change without a code change on your side. API keys created since May 28, 2026 are auth keys by default and unrestricted standard keys are rejected.

On the data side, the paid tier promise is clean: prompts and responses aren’t used to improve products and are logged only briefly for abuse checks. The free tier is the opposite, so the real account risk for a company is a developer testing with customer data on an unbilled project. Project-level spend caps, set on the Spend page, are the easiest guardrail. Use them.

Whether to build on it

For prototyping and low-volume products, yes. No card, free current Flash models, and a $5 entry to the paid tier make it the lowest-friction API we track. For production, it’s worth it if you accept opaque free limits, a prepay model that hard-stops at zero, and prices that step up in December and January. Teams with Google Cloud credits should call Gemini through Vertex AI instead, where the same models are covered by the credits this API no longer accepts.