Google released Gemini 3.7 Flash on August 13 — just three weeks after 3.6 Flash — with meaningful gains on coding and agent benchmarks. But the number that matters most for anyone managing an AI budget isn't the benchmark score: it's the date the promotional price expires.
What changed
According to DataCamp, The Decoder, and OpenRouter, Gemini 3.7 Flash jumped from 34.4% to 43.6% on FrontierCode 1.1 Main, from 49% to 65.3% on DeepSWE v1.1, and hit 1,588 Elo on the Code Arena web-development leaderboard, plus 30.4% on AutomationBench — gains Google attributes to improvements in multi-step planning, tool use, and the ability to adapt to roadblocks mid-task, not just more parameters or training data. The model is available through Antigravity, AI Studio, Vertex AI, and the Gemini Enterprise Agent Platform, covering both individual use and enterprise production.
On pricing, Google did something few competitors do: it publicly announced the expiration date of its own promotion, in the same launch announcement. The launch price — $0.75 per million input tokens and $3.75 per million output tokens — holds through December 31, 2026. Starting January 1, 2027, the price doubles: $1.50 for input and $7.50 for output per million tokens. That's not market speculation — it's a schedule the vendor itself has already published.
Why it matters
Yesterday we showed here how DeepSeek repriced overnight, with little advance notice, citing capacity constraints as the reason. Gemini 3.7 Flash is the direct counterpoint: same promotional pricing, same commercial logic of driving usage while the model is new, but with a public expiration date from day one. For anyone building an AI budget around price per token, the difference between these two pricing-communication models is the number that actually matters over the medium term — not which one is cheaper today.
The impact for Brazil
For Brazilian companies planning 2027 AI budgets around Gemini 3.7 Flash, the math is already there: any project still in production after January 1 pays double per token, and that should go into the financial spreadsheet now, not get discovered on January's invoice. It's a real planning advantage over vendors that reprice without notice — but it only becomes an actual advantage if the company uses that information today, instead of treating the current price as permanent until told otherwise.
Entercast's take
Placing Gemini 3.7 Flash next to this week's DeepSeek repricing makes one vendor-evaluation criterion clear, and it matters more than price itself: does the vendor disclose when a price will change, how much advance notice it gives, and does it do so publicly and verifiably? That doesn't replace the discipline of reviewing actual cost per task periodically, which we've recommended here before — but companies that favor vendors with transparent pricing policies, even if it means paying a bit more today, tend to have fewer budget surprises the following year. That's worth weighing alongside the benchmark score when deciding where to run production workloads.