Gemini 3.5 Flash

Google's official AI for Developers documentation confirms that Gemini 3.5 Flash is generally available, stable, and ready for scaled production use, positioned as Google's most intelligent Flash model with sustained frontier performance in agentic execution, coding, and long-horizon tasks. The stable model ID is `gemi For developers, the page serves as the authoritative migration and quickstart reference: all examples use the Interactions API, with GenerateContent API also supported under the same configuration. Key new capabilities highlighted include sub-agent deployment and rapid agentic loops, iterative coding cycles with rapid

Constellation Research's enterprise-focused coverage of the Google I/O 2026 launch confirms that Google launched Gemini 3.5 Flash as a series designed to combine frontier intelligence with action at reasonable token costs, with Sundar Pichai noting internal use had already shifted to the new Flash model. Pichai is quot The article frames Gemini 3.5 Flash as supporting multi-agent autonomous sessions, complex coding pipelines, iterative research projects, and long-running projects, with rollout across Google's products and APIs. It also notes that Gemini 3.5 Pro is rolling out the following month and will be embedded into Antigravity

The official Google Cloud blog post by CEO Thomas Kurian, summarizing Google I/O 2026 innovations, confirms that the Gemini 3.5 family launched starting with Gemini 3.5 Flash, framed as combining "frontier intelligence with action." The post attributes the models to Google DeepMind, noting they were engineered from the Google Cloud describes Gemini 3.5 Flash as delivering intelligence that rivals large flagship models on multiple dimensions at Flash-tier speeds, calling it the strongest agentic and coding model in the Flash series to date. Distribution is across Google Cloud via Gemini Enterprise, the Gemini Enterprise Agent Platform

Ken Huang's Substack analysis (published May 20, 2026) frames Google I/O 2026 as Google building a "distributed agent runtime" rather than a single product launch, positioning Gemini 3.5 Flash as the reasoning-and-action model at the center of a stack that also includes Antigravity (orchestration harness), Managed Agen The same article describes Google's stated thesis of "the agentic Gemini era," noting that Chrome and WebMCP are positioned to make the web legible to agents, and that Search can use the same model-and-harness combination to build custom interfaces, dashboards, and trackers. Beyond restating that 3.5 Flash is the agent

The llm-stats launch recap reports that Google released Gemini 3.5 Flash at Google I/O 2026 on May 19, 2026, as the first model in the new Gemini 3.5 family, with general availability immediately across the Gemini API, Google AI Studio, Google Antigravity, the Gemini app, and AI Mode in Google Search. The stable API mo The recap states pricing of $1.50 per 1M input tokens and $9.00 per 1M output tokens, with $0.15 cached input, and $1.65/$9.90 in non-global regions, alongside a context window of 1,048,576 input / 65,536 output tokens. Modalities are listed as text, image, audio, and video input with text output; dynamic thinking is o

A third-party guide on Digital Applied reports that Google DeepMind released Gemini 3.5 Flash on May 19, 2026, as the first member of the Gemini 3.5 family and the latest in the 3.x Flash line, with the stable API model ID `gemini-3.5-flash` replacing the earlier `gemini-3-flash-preview` identifier. The guide notes a 1 The same guide claims that on Google's published benchmark table, Gemini 3.5 Flash leads Claude Opus 4.7 and GPT-5.5 on five separate evaluations, and confirms general availability across the Gemini app, AI Mode in Google Search, Google Antigravity, the Gemini API (AI Studio and Android Studio), Gemini Enterprise, and

Google's official blog announced Gemini 3.5 on May 19, 2026, kicking off the family with the release of 3.5 Flash, explicitly introduced by DeepMind CTO Koray Kavukcuoglu alongside Chief Scientist Jeff Dean and VPs Oriol Vinyals and Noam Shazeer. The post states 3.5 Flash delivers frontier performance for agents and co The post frames 3.5 Flash as ideal for long-horizon agentic tasks and notes that Gemini 3.5 Pro was already in internal use at launch and slated to ship the following month. 3.5 Flash lands in the top-right quadrant of the Artificial Analysis intelligence-versus-speed index, positioning Flash-tier latency with flagship

An independent benchmark aggregator profiles Gemini 3.5 Flash with a composite score of 68.1 out of 100, ranking it 25th of 232 tracked models based on 27 source-displayable benchmark rows. The model shows particularly strong category percentile rankings in Knowledge (90th), Multimodal (89th), Coding (79th), and Agenti The profile reports API pricing at $1.50 input and $9 output per million tokens, with cached input at $0.15, batch/cache pricing as low as $0.075, and a blended rate of $5.25. Speed is measured at 210 tok/s with a first-token latency of 17.45 seconds and a 1M token context window, giving developers a third-party snapsh

Appwrite's independent deep dive, published 2026-05-19 at the time of Google I/O, frames Gemini 3.5 Flash as built on the Gemini 3 Flash reasoning foundation with explicit thinking levels controlling quality, cost, and latency, with the headline numbers corresponding to the high thinking configuration. The article reco The piece supplies API pricing of $1.50 per million input tokens, $9.00 per million output tokens, and $0.15 per million cached input tokens, noting this makes Gemini 3.5 Flash the most expensive Flash-tier model Google has released to date. It also publishes a head-to-head benchmark table against Gemini 3 Flash, Gemin

The llm-stats model profile page tracks Gemini 3.5 Flash on a composite LLM Stats Score of 42.6 with a blended price of $1.86 per million tokens, ranking it against Gemma 4 E4B, GPT OSS 120B, DeepSeek-V4-Flash-0731, Muse Spark 1.3, and GPT-6 Astra in cost efficiency comparisons. Quality tracker readings are reported as Benchmark scores on the same page are sourced to deepmind.google and include CharXiv-R at 0.84 (rank 18), MCP Atlas at 0.84 (rank 5), and MMMU-Pro results, all scored on a normalized 0–1 scale. The page is presented as a secondary aggregator of Google's published benchmark numbers rather than a first-party measurement,