AI & Services

Gemini 3.7 Flash Posts a Record First Week.

Google CEO Sundar Pichai said Gemini 3.7 Flash had its biggest first-week growth ever. Here are the ARC Prize verified scores and the official pricing.

· Huuush

Gemini 3.7 Flash Posts a Record First Week

Key takeaways

  1. 1Sundar Pichai said Gemini 3.7 Flash set a record for the biggest growth in its first week after launch
  2. 2ARC Prize verified results are 84.6% on ARC-AGI-2 ($0.25) and 95.5% on ARC-AGI-1 ($0.12), a low-cost, high-score position
  3. 3The current introductory price applies only through December 31, 2026, and doubles starting January 1, 2027
ARC-AGI 2 leaderboard graph with Gemini 3.7 Flash highlighted in yellow

Sundar Pichai, CEO of Google and Alphabet, said on his X account at 12:37 p.m. KST on August 22, 2026 that Gemini 3.7 Flash "broke the record for Gemini growth in its first week after launch." He called it "the fastest-growing model we've built" and added that the model now also powers Google Search and the Gemini app. (Source: X, Sundar Pichai, 2026-08-22)

The ARC-AGI verified scores Pichai cited

The post Pichai quoted is a verified result published by the official account of the benchmark organization ARC Prize at 3:04 a.m. KST on August 21, a day earlier. ARC Prize said "Gemini 3.7 Flash scored 84.6% on ARC-AGI-2 ($0.25 per task) and 95.5% on ARC-AGI-1 ($0.12 per task)," adding that "compared with other frontier models, it sits at a point where low cost and high scores appear together." (Source: X, ARC Prize, 2026-08-21)

In the cost-versus-score chart ARC Prize published, Gemini 3.7 Flash sits alone at a point that reaches a score range similar to top-tier models such as GPT-5.6 Sol and Claude Fable 5 at a much lower cost. Even if its score isn't the highest, it means almost no competing model sits in the same cost range.

The price increase comes January 1

According to the official Google DeepMind page, Gemini 3.7 Flash currently costs $0.75 per million input tokens and $3.75 per million output tokens. This is an introductory price that applies through December 31, 2026, and from January 1, 2027 it rises to $1.50 and $7.50 respectively. (Source: Google DeepMind official page)

The same page also publishes performance against the previous 3.6 Flash. On the software engineering benchmark DeepSWE v1.1 it scored 65.3%, up 16.7 points from 3.6 Flash (48.6%), and on AutomationBench, which measures enterprise workflow automation, it scored 30.4%, up 13.4 points. That is a big improvement for a successor released three weeks later.

Developers who try Gemini 3.7 Flash now still have a window to get a roughly half-price introductory rate along with the latest benchmark scores. But the cost structure changes after the switch to standard pricing on January 1, 2027, so if you plan to put heavy traffic on it, factor that date into your cost calculations first.

Frequently asked questions

When and where can I use Gemini 3.7 Flash?
It's already generally available in the Gemini app, Google AI Studio, the Gemini API, and Google Antigravity, and Sundar Pichai said it has also been applied to Google Search.
Why can I trust the ARC-AGI scores?
Because the results were published by ARC Prize, a third-party benchmark organization, after its own verification process. They aren't figures announced by the company itself but have been cross-checked by an outside body.
What should I check before the switch to standard pricing?
The current introductory price of $0.75 per million input tokens and $3.75 for output is held only through December 31, 2026. From January 1, 2027, it rises to $1.50 and $7.50 respectively, so if you plan to put heavy traffic on it, factor that date into your cost calculations.
How much better is it than 3.6 Flash?
According to Google DeepMind's official announcement, it rose 16.7 points on DeepSWE v1.1 and 13.4 points on AutomationBench. That is a big improvement for a successor released three weeks later.

Related articles