Gemini 3.6 Flash Is Live: Release Date and Availability
gemini-3.6-flash. EvoLink now exposes the same model ID through the Google-native API format; the complete developer guide covers integration, thinking, cost, and controlled rollout.EvoLink availability, checked July 22, 2026:gemini-3.6-flashis callable. The recommended Base URL ishttps://direct.evolink.ai. Standard input and output cost$1.35and$6.75per million tokens—10% below Google's Standard list prices.
Release status at a glance
| Item | Confirmed status |
|---|---|
| Release date | July 21, 2026 |
| Google status | Generally available |
| Public model ID | gemini-3.6-flash |
| Google channels | Gemini API, AI Studio, Android Studio, Antigravity, Gemini Enterprise, Gemini app |
| EvoLink route | Live in Google-native API format under gemini-3.6-flash |
| EvoLink Standard pricing | $1.35 input and $6.75 output per million tokens |
gemini-3.6-flash-tiered label seen in Antigravity was not the documented public API ID.Why “released” does not mean visible everywhere
Three states still need to be distinguished: Google has released the upstream model; individual Google products, accounts, or regions can expose it gradually; and EvoLink has opened its native API route. If the Gemini app does not show the model yet, that does not contradict availability through Google's API or EvoLink.
Compatibility changes developers should notice
| Existing behavior | Gemini 3.6 Flash |
|---|---|
| Sampling controls | temperature, top_p, and top_k are deprecated and ignored |
| Thinking control | Use thinking_level instead of thinking_budget |
| Candidate variants | candidate_count is unsupported |
| Prefilled final model turn | Rejected with HTTP 400 |
| Function responses | Preserve call_id, name, and thought signatures |
The model accepts text, image, video, audio, and PDF input, returns text, supports up to 1,048,576 input tokens and 65,536 output tokens, and defaults to medium thinking.
What developers should do now
- Create a test configuration with the model ID
gemini-3.6-flash. - Remove deprecated or unsupported request fields.
- Confirm account and region availability on the Google surface you use.
- Keep Gemini 3.5 Flash or another proven route as fallback.
- Read the Gemini 3.6 Flash developer guide and validate replay, canary, and fallback before moving production workloads.
- Create an API key in the EvoLink dashboard and send a test request through
https://direct.evolink.ai.
Sources
- Google launch announcement
- Latest-model guide
- Gemini 3.6 Flash model page
- Gemini API pricing
- Gemini API rate limits
FAQ
Is Gemini 3.6 Flash officially released?
Yes. Google made it generally available on July 21, 2026.
What is the official model ID?
gemini-3.6-flash.Why do I not see it in the Gemini app?
Product, account, and regional rollout can lag the upstream API announcement.
Is Gemini 3.6 Flash available through EvoLink?
gemini-3.6-flash through the Google-native API format at the recommended Base URL https://direct.evolink.ai. Standard input and output prices are 10% below Google's Standard list prices.Did request parameters change?
Yes. Sampling parameters are deprecated, thinking control changed, candidate count is unsupported, and a prefilled final model turn is rejected.
