Independent newsroom

Thursday, September 3, 2026

The File

Every claim traces to a primary source. Uncertainty is labeled. Corrections are public.

Labs · Thursday, September 3, 2026

Google introduces Gemini 3.8 Flash and Gemini 3.8 Flash Cyber

The File desk · Sep 3, 2026, 1:42 PM UTC

Status

Confirmed as to what Google published Sept. 2. Benchmark tables are Google’s claims, not independently verified here. No newer Anthropic or Meta frontier primary was fetched overnight.

  • Confirmed

    On Sept. 2, 2026, Google published “Introducing Gemini 3.8 Flash and 3.8 Flash Cyber,” stating Gemini 3.8 Flash’s introductory price is $0.75 per million input tokens and $3.75 per million output tokens, the same as 3.7 Flash, expiring Dec. 31, 2026, then $1.50 / $7.50 starting Jan. 1, 2027, and that Gemini 3.8 Flash Cyber is offered to trusted defenders through a Fairwind Program.

  • Unverified

    Google’s benchmark figures (including HLE-Verified 54.9% and CWE-Bench pass@1 47.2%) are the company’s. This desk did not independently verify them.

On Sept. 2, 2026, Google published “Introducing Gemini 3.8 Flash and 3.8 Flash Cyber.” Gemini 3.8 Flash is described as Google’s most intelligent workhorse Flash model, same introductory price as 3.7 Flash: $0.75 per million input tokens and $3.75 per million output tokens (introductory price expires Dec. 31, 2026; starting Jan. 1, 2027: $1.50 / $7.50). Available to developers (Antigravity, Gemini API / AI Studio / Android Studio / Stitch), Gemini Enterprise, and Google AI Pro/Ultra consumers. Gemini 3.8 Flash Cyber is a cybersecurity variant for trusted defenders via a new Fairwind Program; Google says it ships with more permissive cyber mitigations and is limited to trusted defenders. Company benchmark claims (DeepSWE, CyberGym, CWE-Bench, HLE-Verified 54.9%, and others) are Google’s; not independently verified here. DeepMind’s news hub lists the post under September 2026 Models. Anthropic newsroom still led by Sept. 1 Fable/Mythos 5.1 (no newer frontier post overnight). x.ai/news returned 403. Meta’s AI blog featured Muse Spark 1.1 (July 9, 2026); no matching Sept. 2 Meta primary for a “1.3” release was fetched — this file does not invent a Meta story.

Google’s Sept. 2, 2026, post: “today we’re introducing Gemini 3.8, our best reasoning & coding model yet, at the same speed and low cost of 3.7.” It names two variants. “Gemini 3.8 Flash: our most intelligent workhorse model,” “available at the same introductory price as 3.7 Flash at $0.75 per million input tokens and $3.75 per million output tokens.” Footnote 1: “Introductory price expires on December 31, 2026. Starting January 1, 2027, $1.50/1M input tokens and $7.50/1M output tokens will apply.”

Availability as stated on that page: developers may build in Google Antigravity, or in the Gemini API via Google AI Studio and Android Studio, or generate UIs in Stitch; enterprises may access 3.8 Flash in Gemini Enterprise; consumers on Google AI Pro and Ultra across the Gemini app, AI Mode in Google Search, and Gemini in Google Sheets.

“Gemini 3.8 Flash Cyber: our most capable cybersecurity model with frontier-level performance in vulnerability detection and automated patching, available to trusted defenders through our new Fairwind Program.” Google writes that 3.8 Flash Cyber “ships with a more permissive set of mitigations for cybersecurity, and as such, is only available to trusted defenders.” Fairwind, in the get-started section, is described as providing “trusted government authorities, as well as critical infrastructure operators and software maintainers with prioritized access.” Acceptance criteria beyond that summary were not fetched.

Google’s own benchmark claims on the same page include DeepSWE v1.1, CyberGym, CWE-Bench pass@1 of 47.2% (compared by Google to “a leading frontier model at 47.8%”), and “3.8 Flash also achieves a 54.9% on HLE-Verified.” Those figures are Google’s. This desk did not rerun the benches.

Google DeepMind’s news hub, fetched Sept. 3, lists “Introducing Gemini 3.8 Flash and 3.8 Flash Cyber” under “September 2026 Models.” Anthropic’s newsroom, fetched the same morning, still leads with “Introducing Claude Fable 5.1 and Claude Mythos 5.1,” dated Sept. 1, 2026. openai.com/index returned 403; Wednesday’s File already carried Path to Astra from an archive capture — no new OpenAI post is invented here. x.ai/news returned 403. ai.meta.com/blog returned 200 and featured “Introducing Muse Spark 1.1,” July 9, 2026; no Sept. 2 Meta primary for a newer Muse Spark release was fetched.

What is still unknown or disputed

Primary sources

Every claim in this story is drawn from the documents below. If a fetch failed, that is recorded on the card.

  1. Source 1

    Introducing Gemini 3.8 Flash and 3.8 Flash Cyber

    Google (blog.google) · September 2, 2026

    today we’re introducing Gemini 3.8… Gemini 3.8 Flash… $0.75 per million input tokens and $3.75 per million output tokens… Gemini 3.8 Flash Cyber… Fairwind Program. Introductory price expires on December 31, 2026. Starting January 1, 2027, $1.50/1M input tokens and $7.50/1M output tokens will apply.

    https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/

  2. Source 2

    DeepMind News hub listing

    Google DeepMind · 2026-09

    Introducing Gemini 3.8 Flash and 3.8 Flash Cyber — September 2026 Models.

    https://deepmind.google/blog/

  3. Source 3

    Anthropic Newsroom (checked)

    Anthropic · September 3, 2026

    Top announcement remains Introducing Claude Fable 5.1 and Claude Mythos 5.1 (Sep 1, 2026).

    https://www.anthropic.com/news