Google kicked off its I/O 2026 keynote today with the launch of Gemini 3.5 Flash, the first model in the new Gemini 3.5 family. Described by the company as "the most intelligent model for sustained frontier performance on agentic and coding tasks," it succeeds the Gemini 3 Flash preview released in December 2025 and now becomes the default model across Google's ecosystem — from the Gemini app and AI Mode in Search to the Gemini API, Google AI Studio, Vertex AI, and Android Studio.
What is Gemini 3.5 Flash?
Gemini 3.5 Flash is a general-availability (GA) model that Google says "rivals large flagship models on multiple dimensions" while maintaining the speed and cost-efficiency the Flash series is known for. Its context window supports up to 1 million input tokens and 64,000 output tokens, with a knowledge cutoff of January 2026. Pricing is set at $1.50 per million input tokens and $9.00 per million output tokens — roughly 40% cheaper than Gemini 3.1 Pro and around three times the cost of the previous Gemini 3 Flash, for significantly higher capability.
The model ships with dynamic thinking enabled by default, native function calling, structured output, search-as-a-tool, and code execution. It also introduces support for deploying "teams of subagents," positioning it as a platform for complex multi-step automation.
On benchmarks, Gemini 3.5 Flash posts strong scores across coding and agentic tasks. It scores 76.2% on Terminal-Bench 2.1, 83.6% on MCP Atlas, and 78.4% on OSWorld-Verified — all topping the prior Gemini 3.1 Pro. On reasoning benchmarks it delivers 40.2% on Humanity's Last Exam and 72.1% on ARC-AGI-2. Google emphasizes that the model is 4x faster than comparable frontier models on output generation, measured by tokens per second.
Why It Matters
The launch marks a strategic shift. Google is positioning Gemini 3.5 Flash not just as a faster model, but as the engine for what it calls the "agentic Gemini era." The model is explicitly designed for long-horizon agentic tasks — autonomous workflows that involve planning, tool use, reasoning across extended contexts, and self-correction. This is a direct response to the industry trend toward AI agents that can operate independently over minutes or hours rather than just answering single prompts.
Enterprise integrations were a highlight of the announcement. Shopify is already using Gemini 3.5 Flash with parallel sub-agents for global merchant growth forecasting. Macquarie Bank uses it to reason over 100+ page documents for customer onboarding. Salesforce, Ramp, Xero, and Databricks were also named as early adopters.
For consumers, Gemini 3.5 Flash is rolling out today as the default model in the Gemini app, and in AI Mode in Google Search. Google also used I/O to unveil Gemini Spark, a new personal AI agent that runs on 3.5 Flash and can execute background tasks like parsing credit card statements, monitoring email, and drafting reports — even when the user is offline.
What's Next
Google confirmed that Gemini 3.5 Pro, a larger flagship model, is expected next month. The I/O keynote also introduced Gemini Omni, a multimodal video generation model, and a major Gemini app redesign under the "Neural Expressive" branding. With Gemini now surpassing 900 million monthly active users across 230 countries, the stakes for the 3.5 family are high.
This article is for informational purposes only and does not constitute investment advice. The author may hold positions in securities mentioned. Do your own research before making financial decisions.
