Announced August 13, 2026 — about three weeks after Gemini 3.6 Flash — the new workhorse model posts higher coding and agent scores and lands at an introductory rate roughly half its predecessor’s launch price, while Google’s Gemini 3.5 Pro flagship stays delayed.
2026-08-13 · by Moe Ameen
On August 13, 2026, Google announced Gemini 3.7 Flash, describing it as its most intelligent workhorse model yet for coding and agents. It lands about three weeks after Gemini 3.6 Flash, continuing a fast cadence on Google's cost-efficient Flash tier. This is a text-and-reasoning model aimed at software engineering, web development, and knowledge work — not an image or video generator.
Google reports gains over 3.6 Flash across its benchmarks: FrontierCode 1.1 Main rises to 43.6% (from 34.4%), DeepSWE v1.1 to 65.3% (from 49.0%), WebDev Arena Elo to 1588 (from 1538), a document test it labels GDP.pdf to 34.0% (from 22.0%), and AutomationBench to 30.4% (from 17.0%). Beyond code, the company cites stronger document comprehension for finance, law, and biosciences, plus better tool use, instruction following, and planning for multi-step agent workflows. Google also claimed the model tops rivals such as Claude on some business-workflow tasks — a vendor comparison worth treating as unverified until third parties test it.
The pricing is the sharpest part of the launch. Google set an introductory rate of $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026 — roughly half the previous model's launch price — then $1.50 per million input and $7.50 per million output starting January 1, 2027. The model is available in Google Antigravity, the Gemini API, Google AI Studio, Android Studio, and the Gemini Enterprise Agent Platform, and is rolling into Gemini Spark for Google AI Pro and Ultra subscribers.
The timing carries a subplot: 3.7 Flash shipped while Gemini 3.5 Pro, Google's delayed flagship, still had not — extending that model's wait even as the cheaper Flash line keeps advancing.
Here's the move today, while the introductory price is live: use Gemini 3.7 Flash for exactly the stage it just got cheaper and better at — ideation and planning — and hand the rest to a content engine. A model this cheap is a great place to brainstorm forty hooks, draft a rough script, or outline a month of posts in one agentic pass. But a plan is not a post. The model records no video, designs no carousel, writes no on-brand caption to your voice, and publishes to nothing. That gap did not narrow with this release; it is structural to a language model.
[Kompozy](/) is the layer that closes it. Take the ideas or drafts you generate with 3.7 Flash and drop them in: Kompozy rewrites them in your voice through a [Persona Brief](/glossary/persona-brief), then generates the finished formats — captioned [Persona Shorts](/glossary/persona-shorts) and HeyGen avatar video, brand-exact carousels and quote cards via [HyperFrames](/glossary/hyperframes), photo posts, blog articles, and email newsletters — and schedules and publishes the whole set across the eight social platforms plus blog and email on [Autopilot](/glossary/autopilot). The news is a cheaper brain; the leverage is pairing it with the engine that turns cheap drafts into a published content week.
Google announced it on August 13, 2026, about three weeks after Gemini 3.6 Flash. It is a fast, cost-efficient "workhorse" model for coding, agents, and knowledge work — a text-and-reasoning model, not an image or video generator. Google reports gains over 3.6 Flash on coding and agentic benchmarks.
Google launched it at an introductory rate of $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026 — roughly half the prior model’s launch price — then $1.50 per million input and $7.50 per million output from January 1, 2027. Confirm current rates in the Gemini API docs.
No. It generates and reasons over text — scripts, outlines, captions, plans — but produces no images, video, or scheduled posts. Turning its drafts into finished, on-brand content across platforms is a separate job handled by a content engine like Kompozy.