
AI-generated summary
Google has unveiled Gemini 3.5, starting with the Gemini 3.5 Flash model that promises to outperform Gemini 3.1 Pro in real-world agentic and coding tasks. "3.5 Flash delivers frontier-level intelligence at exceptional speed — proving you no longer have to trade quality for latency."
Announced at Google I/O 2026, this will be Google's default AI model (not to be confused with Flash-Lite), designed to deliver better speed than the current Gemini Pro models at a more affordable price. The tradeoff is lower performance than the 3.5 Pro model (coming next month) in tasks that require deep reasoning and high-context understanding.
However, Google has reduced the compromise between the Pro and Flash models, saying Gemini 3.5 Flash "delivers intelligence that rivals large flagship models on multiple dimensions." It outperforms the current Gemini 3.1 Pro model on coding and agentic benchmarks like Terminal-Bench 2.1 (76.2 percent), MCP Atlas scaled tool use (83.6 percent) and multimodal understanding with 84.2 percent on CharXiv Reasoning. From an output tokens-per-second standpoint, it's four times faster than other frontier models, according to Google.
All of that makes 3.5 Flash ideal for long-horizon agentic tasks, completing what used to take weeks in "a fraction of the time," Google wrote. "Under supervision, it can reliably execute multi-step workflows and coding tasks while sustaining frontier performance." It added that partners, including banks and fintechs, have used it to automate multi-week workflows.
Google noted that 3.5 Flash is now the default model for the Gemini app and for AI Mode in Search worldwide. The personal AI agent Gemini Spark, rolling out to testers today, also runs on 3.5 Flash. At the same time, Google said it strengthened Gemini 3.5's cyber and CBRN (chemical, biological, radiological, nuclear) safeguard so it's less likely to generate harmful content (or mistakenly refuse to answer safe queries). Google detailed the new model's enhanced agentic coding capabilities in a new blog post.
Gemini 3.5 Flash is now "generally available via Google Antigravity, the Gemini API in Google AI Studio and Android Studio, Gemini Enterprise Agent Platform and Gemini Enterprise. It's also now available to everyone in the Gemini app and AI Mode in Search," Google said.
There's no word yet on pricing for paid tiers. As for Gemini 3.5 Pro, Google said: "It's already being used internally, and we look forward to rolling it out next month."

A roundup of tech and science developments including NASA's cancellation of the Swift telescope rescue, Tesla's massive recall in China, Meta's ongoing legal challenges regarding child safety, and the rise of Inner Mongolia as an AI data center hub.

A humanoid robot named Lightning, developed by Honor, completed a 100-metre sprint in 9.32 seconds, beating Usain Bolt's 9.58-second human world record. The feat occurred during preparations for the World Humanoid Robot Games in Beijing.

WIRED reports on Flock Safety's AI tools, OpenAI's safety overhaul, and vulnerabilities allowing 'zombified' expired Visa cards to be used for fraud. Additionally, Apple issued record spyware warnings, and Ukraine targeted Russian retailer Wildberries with cyber and drone attacks.

Ulanqab, Inner Mongolia, is rapidly becoming a major AI data center hub due to low energy costs and cold climate. While companies like DeepSeek and ByteDance invest heavily, the region faces significant water scarcity and ongoing reliance on coal power.

Award-winning Hollywood writers and directors are taking gig jobs paying $12 to $200 an hour to train AI models, seeking income amid a severe industry job slump despite concerns they are helping automate their own professions.

Experts in Silicon Valley warn of an imminent AI takeoff and existential risks, citing recursive self-improvement, autonomous agent breakouts, and a lack of global regulatory coordination amid US-China rivalry.