· CostMyTokens · Analysis · 4 min read
Gemini API Pricing 2026: Gemini 4 Argon vs 3.1 Pro vs 3.8 Flash vs 3.1 Flash-Lite
Gemini API pricing for the four models in our calculator, from Google's official price list — plus how the free tier actually works, and the two price rises already scheduled.
Google’s Gemini API is the only major LLM API with a real free tier, and it is also the one whose price list changes most often. This page lists the official paid prices for the four Gemini text models in our calculator, explains what the free tier does and does not include, and works out what typical monthly workloads cost.
Disclosure: this article contains no affiliate links. Where CostMyTokens does use affiliate links elsewhere, they never affect order — every table is sorted by price only. Read our disclosure.
Gemini API prices (official, per 1M tokens)
Standard tier, sorted by input price. Pro prices are for prompts up to 200K tokens; longer prompts cost more (see the billing rules).
| Model | Input / 1M | Output / 1M | Cached input / 1M | Batch input / output | Free tier | Source |
|---|---|---|---|---|---|---|
| Gemini 3.1 Flash-Lite | $0.25 | $1.50 | $0.025 | $0.125 / $0.75 | Yes | official pricing, verified 2026-10-26 |
| Gemini 3.8 Flash | $0.75 | $3.75 | $0.075 | $0.375 / $1.875 | Yes | official pricing, verified 2026-10-26 |
| Gemini 4 Argon | $2 | $10 | $0.10 | $1 / $5 | No | official pricing, verified 2026-10-26 |
| Gemini 3.1 Pro | $2 | $12 | $0.20 | $1 / $6 | No | official pricing, verified 2026-10-26 |
Two of these prices are temporary. Gemini 4 Argon’s rates are introductory: they run through November 30, 2026, and from December 1 the model costs $4 / $20. Gemini 3.8 Flash is on a promotional rate through December 31, 2026 and doubles to $1.50 / $7.50 on January 1, 2027. Cached and Batch rates move by the same proportions.
The free tier, precisely
Two of the four models are free of charge: Gemini 3.8 Flash and Gemini 3.1 Flash-Lite include free input and output tokens. Gemini 3.1 Pro and Gemini 4 Argon have no free tier at all. What “free” comes with:
- Rate limits are per project, and there is no fixed public table. Limits are measured in requests per minute, tokens per minute and requests per day; Google shows your current limits in AI Studio and notes they are not guaranteed.
- Free-tier content is used to improve Google’s products. On the paid tier it is not. For anything confidential, that difference matters more than the price.
- Moving up is a ladder. Linking a billing account makes a project Tier 1 (billing cap $250). Tier 2 follows after $100 paid plus 3 days (cap $2,000), Tier 3 after $1,000 paid plus 30 days (cap $20,000 and up). Paid projects also get spend-based rate limits per 10 minutes: $10 on Tier 1, $50 on Tier 2, $200 on Tier 3.
What a month costs
Each workload is 80% input and 20% output, at the Standard tier, with Pro at its up-to-200K rates and Argon at its introductory price. Sorted by cost.
| Model | 10M tokens / month | 100M tokens / month | Same, 50% of input cached | Same, via Batch API |
|---|---|---|---|---|
| Gemini 3.1 Flash-Lite | $5.00 | $50.00 | $4.10 (10M) | $2.50 (10M) |
| Gemini 3.8 Flash | $13.50 | $135.00 | $10.80 (10M) | $6.75 (10M) |
| Gemini 4 Argon | $36.00 | $360.00 | $28.40 (10M) | $18.00 (10M) |
| Gemini 3.1 Pro | $40.00 | $400.00 | $32.80 (10M) | $20.00 (10M) |
Note that Pro and Argon share a $2 input price, but Pro’s $12 output makes it the most expensive of the four at this mix — and from December 1, Argon’s row doubles to $72.00 per 10M tokens when its introductory pricing ends.
Billing rules that change the bill
- Two price rises are already scheduled. Argon doubles on December 1, 2026, and 3.8 Flash doubles on January 1, 2027. Any budget built on today’s Argon or Flash prices expires on those dates.
- Pro charges more past 200K tokens. A prompt over 200K tokens is billed at $4 input / $18 output, cached reads cost $0.40, and Batch rises to $2 / $9. Our calculator uses the up-to-200K rates.
- Thinking tokens are billed as output. Every Gemini output price includes thinking tokens, so reasoning-heavy workloads pay the output rate for tokens you never see.
- Explicit caching also charges storage by the hour. Cached reads are cheap (from $0.025 per 1M on Flash-Lite), but explicitly stored context costs extra per 1M tokens per hour: $0.50 on 3.8 Flash, $1.00 on Flash-Lite and $4.50 on 3.1 Pro. Argon is the exception — it supports implicit caching only, so hits are automatic and there is no storage fee.
Which model to price first
- Gemini 3.1 Flash-Lite: the cheapest paid tier, and free to start. High-volume, simple work.
- Gemini 3.8 Flash: the strongest Flash, also free to start — but budget for the January 1 doubling.
- Gemini 4 Argon: the flagship. Its introductory price ends November 30, so comparisons made after that date should use $4 / $20.
- Gemini 3.1 Pro: the previous-generation flagship, with the 200K price step. At an 80/20 mix it currently costs more than Argon despite the identical input price.
Price your own traffic on the Flash-Lite, Flash, Argon or Pro pages. All Google models are listed on the provider page.
Prices: Google Gemini API pricing and rate limits, verified 2026-10-26. Prices change; check the official pages before committing budget.