Google ships Gemini 3 flash everywhere—and makes speed the default
New Capabilities
Eight months in: Gemini 3.5 Flash is generally available and the new default on every major Google surface, AI Mode has one billion monthly users, and Gemini 3.5 Pro is days away.
Eight months in: Gemini 3.5 Flash is generally available and the new default on every major Google surface, AI Mode has one billion monthly users, and Gemini 3.5 Pro is days away.
When Google launched Gemini 3 Flash in December 2025, it bet that a fast, cheap model could become the default brain of Search, the Gemini app, and developer tooling. That bet compounded. Google shipped Gemini 3.1 Pro in February 2026 and Gemini 3.5 Flash in May; the latter reached general availability on July 14 and is now the default across every major Google surface.
AI Mode passed one billion monthly users at I/O 2026 in May, expanding to 200 countries without a subscription requirement. Gemini 3.5 Pro, a full architectural rebuild, is expected within days. The original preview failed at complex SVG layouts and recursive tool-calling; third-party reporting targets July 17 for the release, though Google has not confirmed the date.
Google announced AI Mode crossed one billion monthly users at I/O 2026 in May, one year after its launch as a U.S.-only Labs experiment. Queries have more than doubled every quarter since launch.
$1.50 / $9
Gemini 3.5 Flash input/output cost per 1M tokens
Gemini 3.5 Flash, now the default Flash model, is priced at $1.50 input and $9.00 output per million tokens. That is 3x the price of the original Gemini 3 Flash but the model outperforms Gemini 3.1 Pro on coding and agentic benchmarks while running four times faster.
77.1%
ARC-AGI-2 score (Gemini 3.1 Pro, February 2026)
Google's Gemini 3.1 Pro scored 77.1% on ARC-AGI-2 at launch in February 2026, more than double its predecessor. Google claimed benchmark leadership across 13 of 16 major evaluations at release.
Gemini 3.5 Flash scored 76.2% on Terminal-Bench 2.1 and 83.6% on MCP Atlas agentic benchmarks at I/O 2026, outperforming Gemini 3.1 Pro on both coding and multi-step agent tasks.
Voices
Curated perspectives — historical figures and your fellow readers.
Ever wondered what historical figures would say about today's headlines?
Sign up to generate historical perspectives on this story.
17 events
Latest: 2 days ago
Showing 8 of 17
JK to step
Tap a bar to jump to that date
Jump to
July 2026
Reports target July 17 for Gemini 3.5 Pro after a full architecture rebuild
LatestLaunch
Third-party reporting placed Gemini 3.5 Pro's general availability on July 17, 2026 after Google scrapped its initial architecture. The original preview failed at complex SVG layout generation and recursive tool-calling; no model card, pricing page, or confirmed API endpoint had appeared in public Gemini API documentation as of July 13.
May 2026
AI Mode crosses one billion monthly users at I/O 2026, expands to 200 countries
Product
Google announced AI Mode had passed one billion monthly users a year after its launch, with queries more than doubling every quarter. The company expanded AI Mode to nearly 200 countries across 98 languages without a subscription requirement and set Gemini 3.5 Flash as the new default model globally.
Managed Agents launch in Gemini API; Gemini CLI transitions to Antigravity CLI
Developer
Google introduced Managed Agents—isolated Linux sandbox environments for AI agents, accessible via a single API call—and released Antigravity 2.0, a standalone desktop app that replaced Gemini CLI as Google's primary agentic development platform.
February 2026
Google launches Gemini 3.1 Pro, claiming ARC-AGI-2 benchmark leadership
Launch
Google released Gemini 3.1 Pro with a 77.1% ARC-AGI-2 score—more than double its predecessor—at $2 input / $12 output per million tokens. The model launched in the Gemini API, AI Studio, Vertex AI, Gemini Enterprise, and Gemini CLI.
January 2026
Gemini 3 Search-grounding billing activates as announced
Developer
Google began charging developers for Search grounding on Gemini 3 API calls at $14 per 1,000 search queries, as flagged in December 2025 pricing documentation. Unlike earlier Gemini generations, Gemini 3 bills per internal search query the model generates, not per user prompt.
December 2025
Vertex AI updates Standard PayGo throughput guidance for Gemini model families
Developer
Google Cloud documents baseline throughput tiers for Gemini Flash/Flash-Lite families (with a 30,000 RPM per-model-per-region system limit) and clarifies burst behavior and 429 handling for shared capacity.
Google posts official Gemini API pricing and rate-limit tables for Gemini 3 Flash Preview
Developer
Gemini 3 Flash Preview appears in Gemini API pricing with batch and context-caching details, while the rate-limits documentation adds explicit batch enqueued-token quotas for Flash by usage tier.
Google launches Gemini 3 Flash
Launch
Google releases Gemini 3 Flash as a faster, cheaper model in the Gemini 3 family.
Gemini app switches its default model
Product
Gemini 3 Flash becomes the default experience, replacing the prior Flash generation.
Search AI Mode rolls out Gemini 3 Flash globally
Product
AI Mode defaults to Gemini 3 Flash worldwide; Pro and image tools expand in the U.S.
Gemini 3 Flash lands in CLI and developer tooling
Developer
Google adds Gemini 3 Flash to Gemini CLI and highlights API availability.
OpenAI ships GPT-5.2 amid competitive pressure
Competition
Reuters reports GPT-5.2 launches after an internal “code red” push.
November 2025
Google launches Antigravity for coding agents
Developer
Google introduces Antigravity, an agentic development platform spanning editor, terminal, and browser.
Gemini 3 hits Search on day one
Product
Google introduces Gemini 3 in Search AI Mode for U.S. subscribers.
Gemini 3 Pro arrives in Gemini CLI
Developer
Google integrates Gemini 3 Pro into its terminal-first developer assistant.
Google debuts AI Mode in Labs, using a custom Gemini model.
Historical Context
3 moments from history that rhyme with this story — and how they unfolded.
1 of 3
2024-07-18
OpenAI releases GPT-4o mini as a cheap default-class model
OpenAI introduced GPT-4o mini as a low-cost model aimed at making high-frequency calls practical. The pitch was not “best model,” but “best economics,” enabling parallel calls and larger context at far lower price.
Then
Developers got a clear path to cheaper agent loops and customer-facing chat at scale.
Now
The market normalized the idea that small models can be “default” without feeling second-rate.
Why this matters now
Gemini 3 Flash is Google’s version of the same move—win by being the default everywhere.
2 of 3
2024-05-14 to 2024-07-25
Google introduces Gemini 1.5 Flash to serve fast, high-volume workloads
Google positioned Flash as the speed-and-efficiency line, explicitly built for lower latency and lower serving cost. It then upgraded the free-tier Gemini experience to Flash, training users to accept “Flash” as the normal experience.
Then
Flash became synonymous with responsiveness, not compromise, in Google’s consumer assistant.
Now
Google built the runway for later generations where Flash can inherit near-Pro reasoning.
Why this matters now
Gemini 3 Flash is the payoff: a speed tier that claims Pro-like intelligence.
3 of 3
2024-03-13
Anthropic launches Claude 3 Haiku as the fast, affordable tier
Anthropic released Haiku as its fastest and most affordable Claude 3 model. The focus was throughput and responsiveness for enterprise workflows, not just top-end reasoning.
Then
Claude became easier to deploy in latency-sensitive, high-volume use cases.
Now
The industry’s product strategy shifted toward tiered families where speed models do most work.
Why this matters now
Gemini 3 Flash follows the same industry arc: the speed tier becomes the business tier.