google ships gemini 3.8 flash to general availability; one user says three questions bought a five-hour wait

Google announced two Gemini 3.8 variants on September 2, 2026: Gemini 3.8 Flash, documented as generally available under the stable model ID gemini-3.8-flash, and Gemini 3.8 Flash Cyber, restricted to trusted defenders through the Fairwind Program. Google's [91] release documentation for Gemini 3.8 Flash ↗ describes Flash as ready for production use with a default thinking level of medium. The [88] launch announcement ↗ lists Flash across the Gemini API, Google AI Studio, Android Studio, Google Antigravity, and Gemini Enterprise, and for Google AI Pro and Ultra subscribers in the Gemini app, AI Mode, and Gemini in Google Sheets. The [89] Gemini API model page ↗ marks computer use as Preview and lists audio generation, image generation, and the Live API as unsupported.

by the numbers

Documented limits, from the [89] Gemini API model page ↗ and the [90] DeepMind model card ↗:

Rates from the [92] Agent Platform pricing page ↗, per million tokens:

The same page prices cached input separately.

Google-reported benchmark figures from the [88] launch announcement ↗, which this packet treats as company-reported claims pending independent verification:

• Flash: 54.9% on HLE-Verified
• Flash Cyber: over 70% on an internal 20-language vulnerability benchmark
• Flash Cyber: 47.2% pass@1 on CWE-Bench, against 47.8% for a leading frontier model

what people are saying

The blocks below are attributed paraphrases of each voice's recorded position, drawn from the fact packet and linked to the receipt. They are not verbatim excerpts from the linked pages.

Google, in the [88] launch announcement ↗, on who can reach the cyber variant:

Google states Flash Cyber is available to trusted defenders through the Fairwind Program.

Google, in the same [88] launch announcement ↗, on how its own performance numbers should be read:

Google notes that higher effort can use more tokens, and that the benchmark and real-world performance statements are company-reported claims; chart-level methodology and comparative results require independent review.

Google DeepMind, in the [90] Gemini 3.8 Flash model card ↗, on what the model gets wrong:

The card's known limitations include hallucinations, occasional slowness or timeouts, higher token use at higher effort, and a March 2026 knowledge cutoff with some domains limited to January 2025.

The official Gemini account, in its [93] availability post ↗:

The account said Gemini 3.8 Flash was available starting that day for Pro and Ultra users.

X user @BayasBochi_S, in a [94] post alleging early usage exhaustion ↗:

The post alleges that Gemini 3.8 Flash exhausted daily tokens after three questions and required a five-hour wait.

That last account is an unverified anecdote. It is one user's report, it has not been confirmed, and it does not establish a service-wide limit. The packet records that it aligns directionally with Google's own higher-token-at-higher-effort caveat without verifying anything about product quotas.

claims

• Confirmed — Google announced Gemini 3.8 Flash and Gemini 3.8 Flash Cyber on September 2, 2026.
• Confirmed — Gemini 3.8 Flash is documented as GA with the stable model ID gemini-3.8-flash.
• Confirmed — Gemini 3.8 Flash has a 1,048,576-token input limit and a 65,536-token output limit.
• Confirmed — Global standard pricing holds its lower input and output rates through December 31, 2026 and rises to higher standard rates on January 1, 2027; the exact per-million figures for both periods are in the pricing table above.
• Supported — Gemini 3.8 Flash Cyber is available to trusted defenders through the Fairwind Program, not as the general Gemini API model documented for Flash.
• Supported — Google reports benchmark and internal-evaluation gains, but this packet treats comparative performance as company-reported claims pending independent methodology and result verification.
• Unverified — The bounded discourse pass produced a usage-cost angle: an X user reported exhausting daily tokens after three questions and waiting five hours, attributing the behavior to higher token consumption. This is an unverified anecdotal report, not a product-limit fact.

receipts

• [88] Introducing Gemini 3.8 Flash and 3.8 Flash Cyber ↗ — Google, official announcement
• [89] Gemini 3.8 Flash model page ↗ — Google AI for Developers, developer documentation
• [90] Gemini 3.8 Flash model card ↗ — Google DeepMind, model card
• [91] What's new in Gemini 3.8 Flash ↗ — Google AI for Developers, release documentation
• [92] Agent Platform pricing ↗ — Google Cloud, pricing documentation
• [93] @GeminiApp post on Pro and Ultra availability ↗ — Google Gemini official account
• [94] @BayasBochi_S post alleging usage exhaustion ↗ — X user, unverified community anecdote