Google · API pricing
Gemini 3.6 Flash pricing
Google shipped Gemini 3.6 Flash on July 22, 2026 at $1.50 per 1M input tokens and $7.50 per 1M output. It arrived the same day as two other models that also carry the word “Flash” — and one of them costs 3x less on output. Picking the wrong one is the most expensive mistake available on this launch.
Prices as published by Google on 2026-07-22.
The short answer
- Gemini 3.6 Flash — $1.50 in / $7.50 out per 1M tokens. The general-purpose model of the three.
- Gemini 3.5 Flash-Lite — $0.30 in / $2.50 out. Cheaper on input by 5x and on output by 3x.
- Gemini 3.5 Flash Cyber — no published price. It is not on general sale; access is a limited pilot. See below.
The three models side by side
Confirmed| Model | Input / 1M | Output / 1M | How you get it |
|---|---|---|---|
| Gemini 3.6 Flash | $1.50 | $7.50 | Gemini API, AI Studio, Android Studio, Antigravity, Enterprise Agent Platform, Gemini app |
| Gemini 3.5 Flash-Lite | $0.30 | $2.50 | Same channels, plus rolling out inside Google Search |
| Gemini 3.5 Flash Cyber | Not published | Not published | Limited-access pilot for governments and trusted partners, via CodeMender |
Google's launch post does not state a context window for any of the three. We have left those cells out rather than carry over a number from an earlier generation — if you are sizing a long-context workload, check the model card before you commit.
What it costs at your volume
The default below is a 5:1 input-to-output ratio, which is typical of retrieval and summarisation work. Push the output number up — as agent and reasoning workloads do — and the gap between the two priced models widens fast, because output is where the 3x sits.
Cost calculator
Enter your monthly token volume to see what each tier would cost.
| Tier | Input cost | Output cost | Monthly total |
|---|---|---|---|
| Gemini 3.6 Flash | $1.50 | $1.50 | $3.00 |
| Gemini 3.5 Flash-Lite | $0.30 | $0.50 | $0.80cheapest |
Estimates only — list prices, before any caching discount, batch pricing, or enterprise agreement.
Which one you actually want
Output-heavy work → the spread is the whole decision.
Output costs $7.50 on 3.6 Flash against $2.50 on Flash-Lite. If your workload generates long responses, that 3x is most of your bill, and it dwarfs any input-side saving.
Latency-bound work → Flash-Lite has the published number.
Google cites Artificial Analysis measuring Flash-Lite at 350 output tokens per second. No equivalent throughput figure is published for 3.6 Flash in the launch post, so treat a speed comparison between the two as unmeasured rather than assuming the pricier model is faster.
Migrating off 3.5 Flash → count tokens, not just the rate.
Google cites the Artificial Analysis Index showing 3.6 Flash uses 17% fewer output tokens than 3.5 Flash on the same work. A per-token rate change and a per-token volume change move your bill in opposite directions, so a rate-only comparison will mislead you. Google's post does not restate 3.5 Flash's own price, so we are not publishing that delta as a number.
Security work → Cyber is not a product you can buy today.
Gemini 3.5 Flash Cyber is described as reaching competitive frontier performance on the CyberGym benchmark, but distribution is a limited-access pilot for governments and trusted partners through CodeMender. There is no price and no self-serve path, so it cannot go into a build plan yet.
What this launch says about Gemini 3.5 Pro
Two things worth noting for anyone waiting on the Pro model. Google says 3.5 Pro is still in partner testing, with broad availability coming “as soon as it's ready” — so the widely repeated July 17 launch date has now passed without it. Google also says pre-training on Gemini 4 has begun. We track the confirmed and unconfirmed claims separately on the Gemini 3.5 Pro release date page.
Sources
- Google — Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber (launch post) — retrieved 2026-07-22
Figures on this page last checked against these sources on 2026-07-22. Vendors change pricing and specs without notice — if a number here disagrees with the vendor's own page, trust the vendor.