When will Gemini CLI be discontinued?

Gemini CLI will be discontinued on June 18, 2026. You'll need to migrate to Antigravity CLI before that date. The official migration guide is already available in Google's documentation.

Is Antigravity 2.0 available outside the US?

The CLI and SDK are available globally. The Desktop app is currently launching primarily in the US, with availability in other regions yet to be confirmed. Starting with the CLI is recommended for evaluating workflow compatibility.

Is the Google AI Ultra $100 plan available globally?

AI Ultra is currently targeting the US market primarily. Google One AI Premium (which includes advanced Gemini features) is available in many countries at local pricing, though the feature set differs slightly, and features like Gemini Spark may not be included.

Is Gemini 3.5 Flash stable enough for production use?

As of May 25, 2026, Gemini 3.5 Flash is still in Preview status. For production environments, monitor official announcements closely — there may be API breaking changes after the GA release. Version-locking in production pipelines is strongly recommended.

Google I/O 2026: Your AI Tool Stack Decision Guide

Google I/O 2026 is over. Most coverage tells you "Gemini 3.5 Flash launched" and "the AI Agent era is here," but misses the actual competitive shift: this event changed the dynamics between Claude, GPT, and Cursor. Gemini 3.5 Flash's input token pricing is half that of Sonnet 4.6 (output is 60%). Antigravity 2.0 directly challenges Cursor. The $100 AI Ultra plan significantly cuts individual agentic workflow costs. Based on our cross-referenced analysis of the announcements, you don't need to chase every new model — but you do need to know which announcements actually changed your tool options.

TL;DR

Gemini 3.5 Flash: Intelligence Index 55 (vs Claude Sonnet 4.6 at 52), 280+ tokens/sec, priced at $1.50/$9 per M tokens — excellent for agentic tasks, but still slightly behind Sonnet 4.6 for production code review
Antigravity 2.0: Replaces Gemini CLI with a June 18 deadline, offers Desktop app + CLI + SDK, directly competing with Cursor and Claude Code
Google Search Agent: AI Mode surpasses 1B monthly active users, Information Agent launching this summer, some features still US-only
Google AI Ultra $100: Dropped from $250 to $100, includes Gemini Spark (US-first)
Regional limits: Gemini Spark, AI Inbox, and Daily Brief are currently US-only

Gemini 3.5 Flash: What's Actually Different

This isn't a minor Flash update. Gemini 3.5 Flash is Google's first model combining flagship-level intelligence with Flash-tier speed, scoring 55 on Artificial Analysis's Intelligence Index — above Claude Sonnet 4.6 (52) and below GPT-5.5 (60).

Key specs worth noting:

Speed: 280+ output tokens/sec, roughly 2.1x faster than Claude Sonnet 4.6
Pricing: $1.50/$9 per M tokens (input/output). Sonnet 4.6 is $3/$15. GPT-5.5 is $5/$30.
Context window: 1 million tokens
Agentic tasks: GDPval-AA score of 1,656 Elo, Terminal-Bench 76.2% — leads the Intelligence vs Speed Pareto frontier among current models

Two important caveats. First, for production-grade code review, Claude Sonnet 4.6 is still the safer choice — the coding accuracy gap becomes more pronounced in complex refactoring. Second, this represents a 3x price increase from Gemini 3 Flash ($0.5/$3 to $1.50/$9). If your pipeline had heavy reliance on the old Flash tier, recalculate costs before assuming you're saving money.

Our assessment: Gemini 3.5 Flash fits well as the sub-task layer in agentic pipelines (high-volume, lower-complexity parallel calls), while primary reasoning stays on Sonnet 4.6 or GPT-5.5. It's a cost-reduction complement, not a wholesale replacement.

Antigravity 2.0: 5 Things Developers Need to Know

The Antigravity 2.0 demo at I/O was the most striking thing shown: 93 sub-agents running in parallel, over 15,000 model requests, 2.6 billion tokens consumed, a working OS built in 12 hours (Doom ran), total cost under $1,000.

But before you rush to try it, understand what it actually is:

1. It's an agent orchestration platform, not an IDE

You don't use Antigravity to write code. You use Antigravity to manage agents that write code. Five components: Desktop app (visual parallel agent interface), CLI (terminal-first), SDK (self-hosted), Managed Agents API (spin up a Linux sandbox with a single API call), Enterprise (Google Cloud integration).

2. It's not a replacement for Cursor or Claude Code — yet

Cursor excels at single-file editing and IDE integration. Claude Code shines at autonomous task completion. Antigravity targets multi-agent orchestration and cloud-native deployment, better suited for parallel CI/CD pipelines than daily component refactoring. Short term, these three tools are complementary.

3. Gemini CLI discontinuation deadline: June 18

This is the one item requiring immediate action. If any part of your workflow depends on Gemini CLI, you must complete migration to Antigravity CLI before June 18. CLI and SDK are available globally, and the migration guide is live in official documentation.

4. Desktop app availability outside the US is unconfirmed

The Desktop app is launching primarily in the US. Availability in other markets has not been confirmed. Start with the CLI version to evaluate workflow compatibility.

5. Cost control matters in high-parallelism agentic tasks

The OS demo cost under $1,000, but ran 2.6B tokens over 12 hours. At scale, multi-agent workflows can accumulate significant API costs quickly. Set budget caps before running complex agentic tasks.

For context on the broader CLI landscape, see: Claude Code vs Gemini CLI vs Codex CLI Decision Guide

Google Search Agent: The End of Information Anxiety, or the Start of a New Kind?

The Search changes this time are more significant than the model releases. AI Mode reached 1 billion monthly active users within a year — the biggest functional overhaul to Google Search in 25 years. New capabilities include uploading images, videos, files, and Chrome tabs directly into search. Gemini 3.5 Flash is now the global default model for AI Mode.

Two new features to understand:

Information Agent: Monitors topics you specify 24/7, proactively pushing relevant updates. Think of it as "a permanently running search agent" — housing markets, specific stocks, industry news. Set the conditions and it notifies you automatically. Launching this summer, US AI Pro/Ultra users first.

Booking Agent: Can make phone calls to place reservations on your behalf. The practical value in non-US markets depends heavily on local service infrastructure and language logistics.

For users outside the US: Personal Intelligence has expanded to nearly 200 countries and 98 languages, and Gemini 3.5 Flash as the AI Mode default is live globally. But Information Agent and Booking Agent are US-first for summer launch. Watch for regional expansion announcements.

Practical advice for product managers: Don't wait for Information Agent to start using Search for competitive monitoring. Today's AI Mode already supports complex multi-part queries and multimedia input. Build manual "competitive monitoring" search templates now and run them regularly. That's a workable bridge while you wait.

Gemini Spark and Workspace AI: Has the 24/7 AI Assistant Actually Arrived?

Gemini Spark is the announcement that caught the attention of many PMs and knowledge workers: delegate tasks through a dedicated Gmail address, runs on Google Cloud isolated VMs, executes long-running tasks 24/7 even when your computer is off.

Integration scope is broad: the entire Google Workspace suite, plus Microsoft SharePoint/OneDrive, ServiceNow, Canva, OpenTable, and Instacart. In theory, you can delegate "monitor inbox, draft replies, update project timeline" to Spark and have it run autonomously around the clock.

Google Workspace also got several updates this cycle:

Google Pics: AI image generation and editing within Slides and Docs — move objects, change text, translate text within images
Gmail Live / Docs Live: Voice input assistant that transcribes and auto-formats
AI Inbox: Smart priority sorting for email plus draft suggestions

The current reality outside the US: Gemini Spark is currently US-only for AI Ultra subscribers ($100/month). AI Inbox and Daily Brief are also US-only. Google Workspace AI Ultra for enterprise is priced separately from personal plans. Timeline for Google Pics and voice features in other regions is unconfirmed.

In short, most of the Workspace AI updates announced at I/O are in "watch and wait" status. If you're already subscribed to Google One AI Premium at local pricing, continue — and wait for official announcements about regional feature expansion.

Tool Decision Framework: Switch or Stay?

Based on our evaluation, here's a decision matrix by role:

Role	Primary Need	Recommended Action	Rationale
Frontend/Full-stack engineer	Code review accuracy	Claude Sonnet 4.6 (primary) + Gemini 3.5 Flash API (for agentic cost reduction)	Sonnet coding accuracy edge matters in complex work; Flash reduces cost for high-volume agentic calls
Full-stack / DevOps	Agent orchestration, CI pipelines	Antigravity 2.0 worth evaluating (start with CLI)	Managed Agents API lowers infra barrier; monitor CLI stability post-6/18
Designer	Image generation/editing	Wait for Google Pics regional availability; use Adobe Firefly or Midjourney for now	Google Pics functionality is compelling but timeline is uncertain
Product Manager	Email/doc workflow automation	Hold on Gemini Spark; test current Workspace AI with Google AI Pro $20	Spark is promising but regional timeline unknown
Everyone	Subscription value	Google AI Ultra $100 vs Claude Max $100 vs ChatGPT Plus $20	Depends on your ecosystem: Google-heavy users consider Ultra; code-heavy work consider Claude Max

On subscription tiers: Google AI Ultra dropping from $250 to $100 is a meaningful pricing signal. In the US, it now sits at the same price point as Claude Max. For users outside the US, local Google One AI Premium pricing applies, with a somewhat different feature set. Until Gemini Spark and related features confirm regional availability, Claude Max $100 remains the more predictable investment for code-focused workflows.

Timeline and Action Checklist

Act now (this week):

Gemini CLI users: Check your pipeline dependencies and complete Antigravity CLI migration before June 18
Try Gemini 3.5 Flash API free in Google AI Studio — identify which sub-tasks in your agentic pipeline could switch to reduce costs
Confirm AI Mode is active on your Google account (most accounts have it enabled globally)

Wait until summer:

Information Agent (Google announced summer launch, US-first)
Google Pics regional availability
Gemini Spark regional expansion

Ongoing evaluation:

Whether Google AI Ultra $100 displaces Claude Max $100 (dependent on Gemini Spark regional rollout)
Whether Antigravity 2.0 affects your Cursor subscription decision (assess post-6/18 CLI stability)

Conclusion: Two Things Actually Matter Right Now

The most important signal from Google I/O 2026 is that Google played both the cost card and the ecosystem integration card at once. Gemini 3.5 Flash compresses API costs for agentic workflows. Antigravity 2.0 plus Google Workspace AI is beginning to form a competitive path distinct from Anthropic and OpenAI.

For digital workers, the two things worth doing right now:

Add Gemini 3.5 Flash to your API tool evaluation list, especially for sub-task cost optimization in agentic pipelines
If you use Gemini CLI, migrate before June 18 — this is the only item with a hard deadline

Gemini Spark, Google Pics, Information Agent? Wait for summer updates and regional availability announcements. Chasing new features isn't the highest-value use of your time right now. Getting more out of your current workflow is.

Where does your current AI tool stack need the most adjustment? If you're using Claude Code for code review and Cursor for daily development, the direct impact of this Google I/O on your workflow is actually limited. Run the Gemini 3.5 Flash API cost calculation, then wait for Antigravity 2.0's stable release before deciding whether to make changes.