Gemini 3.7 Flash: Three Weeks After 3.6, at Half the Price

What Gemini 3.7 Flash is

Gemini 3.7 Flash arrived on 13 August 2026, three weeks after Gemini 3.6 Flash — a gap Google acknowledge directly, attributing it to developer feedback and algorithmic changes they intended to carry into later models. It is the workhorse of the line: a 1,048,576 token context window, 65,536 maximum output tokens, and native input across text, image, video, audio and PDF. Thinking is tunable across low, medium and high, with the minimal level unsupported. The headline of the release was not capability alone but the combination — Google shipped it at half the per-million cost of the model it replaced, three weeks later.

Speed

Gemini 3.7 Flash generates 389.5 tokens per second, with a time to first token of 13.2 seconds. On a measured task it produces around 36,847 output tokens in roughly 90 seconds.

What Google measured against 3.6

Because the two releases are three weeks apart, Google's comparisons are unusually direct, and all of the figures below are theirs rather than independent. On GDP.pdf, which tests processing of complex documents, 3.7 Flash scores 34.0% against 3.6 Flash's 22.0%. On AutomationBench, covering real-world business workflows, 30.4% against 17.0%. On coding, FrontierCode 1.1 Main reaches 43.6% against 34.4%, and DeepSWE v1.1 65.3% against 49.0%. On LMArena's text leaderboard Gemini 3.7 Flash has 1490 Elo (rank 17, high effort); by category, 1520 Elo in Coding (rank 38, high effort), 1503 Elo in Math (rank 18, high effort), 1518 Elo in Expert (rank 21, high effort) and 1490 Elo in Creative Writing (rank 6, high effort). Independently, GPQA Diamond records 95, mock AIME 97 and SimpleQA Verified 69.

Pricing

Input is $0.75 per million tokens and output $3.75, introductory through 31 December 2026, rising to $1.50 and $7.50 on 1 January 2027. Cached input is $0.075 per million through 2026 and $0.15 after, with context storage at $0.50 per million tokens per hour. The Batch API halves the rate to $0.375 and $1.875. There is no long-context surcharge — the flat rate holds across the full window — and unlike most models here there is a genuine free tier, with model and feature limits. It supports web search grounded in Google Search and Maps, code execution, function calling and structured output, and runs on Vertex AI and the Gemini Enterprise Agent Platform. Fine-tuning is not offered. Gemini 3.8 Flash is offered at the same introductory rate.

Availability and safeguards

Google made Gemini 3.7 Flash available to developers in Google Antigravity and the Gemini API through Google AI Studio and Android Studio, to enterprises through the Gemini Enterprise Agent Platform and the Gemini Enterprise app, and to Google AI Pro and Ultra subscribers through Gemini Spark, the 24/7 personal agent in the Gemini app, available in more than 160 countries. Google state it shipped with updated safeguards against misuse in chemical, biological, radiological and nuclear domains and cyber offense, with details in its model card. The model identifier is gemini-3.7-flash. Search grounding includes 5,000 free requests per month shared across Gemini 3.x models, then $14 per 1,000.

Official Resources