Google’s element-themed AI rollout is moving faster than almost anyone expected. Just one day after references to a mysterious new model began circulating in web app code, Gemini 4 Neon has officially surfaced inside Google Cloud’s billing catalog—revealing its exact API pricing and confirming its position as the missing middle tier in Google’s rapidly expanding Gemini 4 lineup.

For developers and enterprise teams watching their token budgets closely, the numbers tell a clear story. Neon is priced at $3 per million input tokens, $15 per million output tokens, and $0.15 per million cached input tokens. That places it squarely between the ultra-fast Hydrogen and Google’s heavyweight flagships, Carbon and Argon—undercutting standard frontier pricing by a flat 25%.

With billing SKUs already populated in Google Cloud and release activity accelerating today, November 18, here is a complete breakdown of how Gemini 4 Neon fits into Google’s pricing ladder and what it means for real-world workloads.

Inside Google’s New Three-Tier Gemini 4 Pricing Ladder

Until now, teams building on Google’s latest generation faced a noticeable gap. On the entry end sat Gemini 4 Hydrogen, engineered for rapid execution and high-frequency everyday tasks. At the top sat Gemini 4 Carbon and Argon, built for deep, long-horizon reasoning and complex software engineering.

The billing catalog entries for Gemini 4 Neon close that gap completely. Starting December 1, when introductory promotional credits on Carbon and Argon expire, Google’s standard global pay-as-you-go lineup shakes out into three distinct tiers:

Model TierPrimary RoleInput (per 1M tokens)Output (per 1M tokens)Cached Input (per 1M tokens)
Gemini 4 HydrogenFast everyday model & high-volume workflows$1.50$7.50$0.075
Gemini 4 NeonBalanced mid-tier intelligence & orchestration$3.00$15.00$0.15
Gemini 4 Carbon / ArgonFrontier reasoning, coding & cyber defense$4.00 (standard)$20.00 (standard)$0.20 (standard)

The math is straightforward. Neon costs twice as much per token as Hydrogen, yet saves developers 25% compared to the $4 / $20 standard rate on Carbon and Argon. Furthermore, leaked billing entries point to Batch and Flex processing tiers dropping Neon’s cost down to $1.50 per million input tokens and $7.50 per million output tokens, giving asynchronous pipelines an even steeper discount.

From Code Leak to Cloud Billing in 24 Hours

The arrival of Neon’s billing SKUs caps off a rapid sequence of sightings. Earlier this week, we covered how Gemini 4 Neon was first spotted in Google’s web app code, where it appeared as an inactive option alongside hints of Google’s element-based naming strategy.

At the time, industry watchers speculated whether Neon would land below Hydrogen as a lightweight “Flash-Lite” successor or step up as a mid-range contender. The Google Cloud billing catalog settles that debate. Rather than competing on rock-bottom latency, Neon is positioned as a true middleweight workhorse—more capable than Hydrogen on multi-step reasoning, without forcing developers to pay full flagship rates on every API call.

Interestingly, recent build references have also linked Neon to orchestrator workflows inside Google’s agentic tooling. In multi-agent setups—where one model plans a complex task and another executes code or tool calls—having a $3 / $15 model bridges the reliability gap when Hydrogen feels slightly underpowered and Carbon feels like expensive overkill.

How Neon Compares to Carbon, Hydrogen, and the Wider Lineup

To appreciate why Google is introducing Neon right now, it helps to look at how quickly the broader ecosystem has evolved over the past few weeks.

When Google rolled out its flagship reasoning powerhouse—detailed in our full breakdown of Google Gemini 4 Carbon’s cybersecurity and long-horizon capabilities—it offered a temporary 50% introductory credit that brought effective costs down to $2 / $10 through November 30. Once December 1 arrives, however, Carbon and Argon jump to their permanent $4 / $20 pricing. Without a middle tier, teams migrating off introductory Carbon pricing would have faced an abrupt 100% cost hike overnight.

Meanwhile, on the speed-first side of the spectrum, Google has been aggressively expanding Hydrogen across consumer and enterprise products—including how Gemini 4 Hydrogen rolled out globally in Google Search to power real-time visual answers and multi-step query fan-out. For a deeper look at how those two ends of the spectrum stack up on raw speed and intelligence scores, check out our Gemini 4 Carbon and Hydrogen benchmarks and pricing comparison.

All of this also ties back to Google DeepMind’s broader periodic-table roadmap. As we explored in our look at how Google’s element-themed Gemini 4 lineup is conceived, assigning distinct elements to specific compute weights gives developers surgical control over latency, reasoning depth, and cost.

Why the 95% Cached Input Discount Matters

Beyond headline input and output rates, Neon’s $0.15 per million cached input tokens rate preserves one of the Gemini 4 family’s strongest financial advantages: a 95% discount on repeated context.

For applications built around massive system prompts, large codebases, or recurring legal and financial documents, context caching dramatically changes the unit economics. Instead of paying $3.00 every time a 100,000-token reference manual is sent with a user query, cached reads cost pennies on the dollar. Combined with Neon’s $15 output rate, teams running high-context retrieval workflows or customer-facing coding assistants get substantial breathing room before hitting frontier spend thresholds.

What to Watch Next

With pricing entries live in Google Cloud’s catalog and release momentum building today, developers should watch for Neon’s official model card, context window confirmation, and thinking-level controls to appear in Google AI Studio and Vertex AI. If Neon mirrors Hydrogen’s granular thinking tiers while offering reasoning closer to Argon and Carbon, it could quickly become the default production choice for teams looking to balance intelligence and API spend heading into 2027.

Add NPowerUser as a preferred source on Google News
Add NPowerUser as a preferred source on Google News