Google Rolls Out Gemini 3.7 Flash Across Search AI Mode, Workspace, Spark and GitHub Copilot

On August 14–15, Google rolled out Gemini 3.7 Flash to Google AI Pro and Ultra subscribers in one concentrated push: it covers Gemini App chat, Google Search AI Mode (English, global), and Google Workspace (debuting with Sheets Canvas); the 24/7 personal AI agent Gemini Spark switched to the model the same day, and the model also landed on GitHub Copilot, gradually opening to all users. The official line: the new model cuts token costs by 50% while keeping low latency, with stronger reasoning and tool-calling, and its DeepSWE v1.1 software engineering benchmark score rose from 37.0% to 65.3%. This is Google pushing the same Flash base across both its own surfaces and third-party developer tools at once, and its cost/performance deserves continued attention.

Confirmed

• Platform coverage: Gemini 3.7 Flash is available to Google AI Pro and Ultra users across Gemini App chat, Google Search AI Mode (English, live globally from today), and Google Workspace, where the first integration is Google Sheets Canvas
• Capability gains: multiple authors relay that the new model does better on instruction following and intent understanding; reasoning accuracy improves on multi-step tasks such as cross-document synthesis; tool-calling with Google Workspace apps is stronger, which the company says pushes personal AI Agent capability further
• Cost and performance: token costs cut by 50% while maintaining low latency; at launch Chandu Thota cited software engineering performance (DeepSWE v1.1) improving from 37.0% to 65.3%, with reasoning in coding and agent loops boosted in step
• Gemini Spark switch: TawohAwa relayed Google's official word that from August 14 the 24/7 personal AI agent Gemini Spark for Pro/Ultra users runs on Gemini 3.7 Flash, handling multi-step tasks with improved accuracy and tool use
• Landing on GitHub Copilot: Gemini 3.7 Flash has launched on GitHub Copilot and is gradually opening to all users; early testing shows improvements in web and app development, agentic coding workflows, code quality, final output presentation, and codebase research

Not Yet Confirmed

• Chandu Thota's post calls the new model "Gemini 3.5 Flash," inconsistent with the "3.7 Flash" wording in other posts—likely a typo; the specs he cited (50% token cost cut, DeepSWE v1.1 from 37.0% to 65.3%) match the other posts

Why It Matters

• The Flash line targets low cost and low latency; another 50% cost cut plus notably stronger reasoning and agent capability means the cost of scaling multi-step Agent tasks keeps falling, and the coding benchmark jump (37.0% to 65.3%) directly improves third-party developer tooling
• Search AI Mode, Workspace, Spark, and GitHub Copilot switched to the same model within two days, showing how fast Google is pushing one base model into both its own ecosystem and external platforms while lifting the value of Pro/Ultra subscriptions

2026-08-14 ~ 2026-08-15 · 9 related posts

Full story(4 episodes)→

• Episode 1: Gemini 3.7 Flash Benchmarks Leak: High Cost-Performance but Instruction-Following Lags(2026-08-13, 90 posts)
• Episode 2: Google Rolls Out Gemini 3.7 Flash Across Search AI Mode, Workspace, Spark and GitHub Copilot(2026-08-14, 9 posts)
• Episode 3: Gemini 3.7 Flash's Tiered Pricing Draws Developer Criticism(2026-08-14, 2 posts)
• Episode 4: Devin adds Gemini 3.7 Flash, matching Claude Sonnet 5 at half the cost(2026-08-14, 3 posts)

• [source] Google Launches Gemini 3.5 Flash: 50% Lower Token Costs, Big Coding Gains — ChanduThota · 2026-08-14
• Google's Gemini 3.7 Flash Now Available in GitHub Copilot, Boosting Agentic Coding — intellectronica · 2026-08-14
• [source] Google: Gemini Spark Now Runs on Gemini 3.7 Flash for Pro/Ultra Subscribers — TawohAwa · 2026-08-14
• Gemini 3.7 Flash launches with price cut, speed gains, and better coding — gerardsans · 2026-08-14
• [source] Google integrates Gemini 3.7 Flash into Search, rolling out in AI Mode — rmstein · 2026-08-15
• Gemini 3.7 Flash Launches with Enhanced Reasoning and Tool Use — arvind_io · 2026-08-15
• Gemini 3.7 Flash Launches with 50% Lower Cost and Boosted Reasoning — gaganghotra_ · 2026-08-15
• Gemini 3.7 Flash expands to Pro and Ultra users across platforms — GeminiApp · 2026-08-15

1 near-duplicate retellings: gaganghotra_