Key findings
- Gemini 3.6 and 3.8 Flash share an introductory price of US$0.75 input and US$3.75 output per million tokens through December 31, 2026, then US$1.50 and US$7.50.
- Flash-Lite list prices rose with each generation, from US$0.10 input and US$0.40 output on 2.5 Flash-Lite to US$0.25 and US$1.50 on 3.1 and US$0.30 and US$2.50 on 3.5.
- Batch and Flex cost half the standard rate. Cached input on 3.8 Flash costs a tenth of the input price, plus an hourly storage charge.
- Priority costs 1.8x the standard rate on 3.8 Flash. One builder measured no latency gain over Standard for text, image, or video in April 2026.
- Thinking tokens bill as output, and turning thinking down can cost accuracy. One builder's PDF-extraction evaluation lost 6% to 20% accuracy on newer models at minimal thinking.
Budget Gemini 3.8 Flash at its standard US$1.50 and US$7.50 from January 1, 2027, twice the current introductory price. Measure how much of your traffic can run on Batch, Flex, or the cache, since each cuts the rate by half or more, and plan on cost per completed task.
Flash list prices by generation
Paid-tier prices in US dollars per million input and output tokens for text, introductory where Google lists one and standard otherwise, checked October 12, 2026 on Google's pricing pages. Gemini 2.0 Flash, shut down on June 1, 2026, listed at US$0.10 and US$0.40.
- Gemini 2.5 Flash: US$0.30 and US$2.50.
- Gemini 3 Flash, pre-release: US$0.50 and US$3.00.
- Gemini 3.5 Flash, as listed on the Agent Platform: US$1.50 and US$9.00; the Developer API routes 3.5 Flash requests to 3.6 Flash.
- Gemini 3.6 and 3.8 Flash: US$0.75 and US$3.75 through December 31, 2026, then US$1.50 and US$7.50. Since October 8, 2026 the Developer API routes gemini-3.7-flash requests to gemini-3.8-flash.
- Gemini 3.5 Flash-Lite: US$0.30 and US$2.50; Gemini 3.1 Flash-Lite: US$0.25 and US$1.50.
Discounts that change the bill
Batch halves the standard rate with a 24-hour target turnaround. The pre-release Flex tier offers the same discount for synchronous requests with best-effort availability. Cached input on 3.8 Flash costs US$0.075 per million tokens through December 31, 2026. Context caching storage costs US$0.50 per million tokens per hour.
One builder moved a document tool from Mistral's OCR API to Gemini Flash at the same inline price and pays half through Batch. Another tagged 260 hours of video for US$11.94 on Flash-Lite.
Where cost estimates go wrong
Thinking tokens bill as output. Image token counts changed between Flash generations, and retired versions can route to successors at new prices. Log tokens per call and the model that answered to see the actual cost.
Teams also report the cost of switching. Several kept Gemini 2.0 or 2.5 Flash-Lite because newer models cost more for work the older ones already did well, and moved to other providers when those models retired.
- Record input, cached, thinking, and output tokens per completed task.
- Price the batch-eligible share at the batch rate.
- Re-run the comparison when a model is routed to a successor or a price period ends.
Methodology
GILD compared Google's Gemini Developer API and Agent Platform pricing pages, release notes, and deprecations page, checked October 12, 2026, and rechecked the release notes on October 25, 2026. Prices are paid-tier prices in US dollars per million tokens for text input, introductory where Google lists one, then standard. Every latency, accuracy, and cost figure from a team or an independent benchmark is that publisher's own measurement, linked in Sources.
Products in this piece
Sources
- Official documentation
- 8
- Public source
- 8
- Gemini API docs: Gemini Developer API pricing
- Google Cloud: Agent Platform pricing
- Gemini API docs: Release notes
- Gemini API docs: Gemini deprecations
- Gemini API docs: Context caching
- Gemini API docs: Batch API
- Gemini API docs: Flex inference
- Gemini API docs: Priority inference
- Hacker News comment by Simon Willison on Gemini 2.0 Flash pricing (Feb 5, 2025)
- Hacker News comment by Simon Willison on Gemini 2.5 Flash pricing (Jul 3, 2025)
- Hacker News comment on Flash-Lite price increases (Jul 21, 2026)
- Hacker News comment on BigQuery workflows on Flash-Lite (Jul 10, 2026)
- Hacker News comment benchmarking the Priority tier (Apr 16, 2026)
- Google AI Developers Forum: An evaluation of 20 PDFs on the pre-release Gemini 3 Flash and newer models (Sep 14, 2026)
- Hacker News comment on moving a document tool from Mistral OCR to Gemini Flash and Batch (Sep 8, 2026)
- Hacker News comment on tagging 260 hours of video with Gemini 3.1 Flash-Lite (May 26, 2026)