Gemini 4 Carbon, Neon, or Hydrogen? API Pricing, Free Tier, and How They Differ from Argon (2026)

What Are Gemini 4 Carbon, Neon, and Hydrogen: Three Releases in November
Google's official changelog records three models reaching GA in November:
- 2026-11-10, Gemini 4 Carbon: the "most capable reasoning model," built for sustained agentic coding, and also strong at multimodal knowledge work, scientific reasoning, and cybersecurity defense
- 2026-11-12, Gemini 4 Hydrogen: "brings cutting-edge intelligence to everyone," with speed and efficiency for everyday coding and agentic workflows; it also became the default model for the Antigravity agent in Gemini Managed Agents
- 2026-11-18, Gemini 4 Neon: frontier-level intelligence with "fast end-to-end task completion" across complex software engineering, knowledge work, and science workflows
Gemini 4 Argon, which reached GA on October 13, was the flagship at the time. The official models page now gives the "strongest reasoning" spot to Carbon and describes Argon as "our efficient Gemini 4 frontier reasoning model," with faster responses and lower token spend per turn than Carbon across complex coding, agentic, math, and science tasks. Argon is not deprecated and has no shutdown date.
Current Gemini 4 Lineup and Pricing
| Model | Model ID | GA date | Paid price (input/output, per million tokens) | Free tier | Official positioning |
|---|---|---|---|---|---|
| Gemini 4 Carbon | gemini-4-carbon | 2026-11-10 | $2/$10 (through 2026-11-30); $4/$20 from 2026-12-01 | No | Strongest reasoning: complex software engineering, multimodal knowledge work, scientific reasoning, cybersecurity defense |
| Gemini 4 Neon | gemini-4-neon | 2026-11-18 | $3/$15 | No | Frontier-level, fast end-to-end task completion |
| Gemini 4 Hydrogen | gemini-4-hydrogen | 2026-11-12 | $1.50/$7.50 | Yes | Speed and efficiency for everyday coding and agentic work |
| Gemini 4 Argon | gemini-4-argon | 2026-10-13 | $2/$10 (through 2026-11-30); $4/$20 from 2026-12-01 | No | Frontier reasoning that is faster and spends fewer tokens per turn than Carbon |
Sources: Gemini API official models, pricing, and changelog pages (aistudio.google.com/docs, checked 2026-11-19), USD, Standard billing. Output prices include thinking tokens.
Pricing: Carbon and Argon Have Two Phases, Neon and Hydrogen Are Flat
All four models publish Standard, Batch, Flex, and Priority prices. Batch and Flex are half of Standard; Priority is 1.8 times Standard.
| Model and billing mode | Input | Output (incl. thinking) | Caching (implicit) |
|---|---|---|---|
| Carbon Standard (through 2026-11-30 → from 2026-12-01) | $2.00 → $4.00 | $10.00 → $20.00 | $0.10 → $0.20 |
| Carbon Batch/Flex (same two phases) | $1.00 → $2.00 | $5.00 → $10.00 | $0.05 → $0.10 |
| Carbon Priority (same two phases) | $3.60 → $7.20 | $18.00 → $36.00 | $0.18 → $0.36 |
| Neon Standard | $3.00 | $15.00 | $0.15 |
| Neon Batch/Flex | $1.50 | $7.50 | $0.075 |
| Neon Priority | $5.40 | $27.00 | $0.27 |
| Hydrogen Standard | $1.50 | $7.50 | $0.075 |
| Hydrogen Batch/Flex | $0.75 | $3.75 | $0.0375 |
| Hydrogen Priority | $2.70 | $13.50 | $0.135 |
Argon's prices and dates are identical to Carbon's in all four billing modes (details in the Gemini 4 Argon guide).
A few things that are easy to miss:
- Carbon and Neon swap places at the end of November: through 2026-11-30, Carbon ($2/$10) is cheaper than Neon ($3/$15); from 12-01, Carbon becomes $4/$20 and costs more than Neon
- Caching is implicit only: all four state "implicit caching only (explicit caching is not supported)," with no separate storage fee
- Hydrogen's free tier has a trade-off: free-tier content is used to improve Google's products, while paid-tier content is not; Batch and Flex are not available on the free tier
- Google Search grounding is billed separately: 5,000 free requests per month (shared across Gemini 3 and newer models), then $14 per 1,000
- Contributor tier: the Hydrogen and Argon price tables also list a Contributor tier, offered as an invite-only private preview, with prices below Standard but with content used to improve products; Carbon has no such tier
Specs: How the Four Differ
| Item | Carbon | Neon | Hydrogen | Argon |
|---|---|---|---|---|
| Context | 1M tokens | 1M tokens (described as a "unified" context) | 1M tokens (described as a "unified" context) | 1M tokens |
| Output per request | 64K by default, up to 1M with long decode | 64K by default, up to 1M with long decode | 64K by default, up to 1M with long decode | 64K by default, up to 1M with long decode |
| Input | Text, images, audio, video, PDF | Same | Same | Same |
| Thinking level | low/medium (default)/high | low/medium/high (default)/xhigh | low/medium/high (default)/xhigh | low/medium (default)/high |
| Caching | Implicit only | Implicit only | Implicit only | Implicit only |
| Free tier | No | No | Yes | No |
| Price changes | Doubles from 2026-12-01 | Flat | Flat | Doubles from 2026-12-01 |
xhigh is the fourth thinking level introduced with Hydrogen, and Neon supports it too. Both default to high, unlike Carbon's and Argon's default of medium. Output prices include thinking tokens, so higher levels usually mean more output tokens; Google's explanation of thinking levels is on the Thinking page at aistudio.google.com/docs.
Cost Math (Arithmetic, Not Benchmarks)
Assumptions: 50 million input tokens and 10 million output tokens per month (thinking included), no caching, Standard billing. The four models think differently, so the same task will not produce the same number of output tokens on each; this compares unit prices only.
| Model and period | Input cost | Output cost | Monthly total |
|---|---|---|---|
| Carbon/Argon (through 2026-11-30) | 50M × $2 = $100 | 10M × $10 = $100 | $200 |
| Carbon/Argon (from 2026-12-01) | 50M × $4 = $200 | 10M × $20 = $200 | $400 |
| Neon | 50M × $3 = $150 | 10M × $15 = $150 | $300 |
| Hydrogen | 50M × $1.50 = $75 | 10M × $7.50 = $75 | $150 |
| (Reference) 3.8 Flash (through 2026-12-31) | 50M × $0.75 = $37.50 | 10M × $3.75 = $37.50 | $75 |
| (Reference) 3.8 Flash (from 2027-01-01) | 50M × $1.50 = $75 | 10M × $7.50 = $75 | $150 |
| (Reference) 3.5 Flash-Lite | 50M × $0.30 = $15 | 10M × $2.50 = $25 | $40 |
The same volume on Batch: Carbon and Argon $100 during the introductory period and $200 from 12-01; Neon $150; Hydrogen $75.
If 40 million of the 50 million input tokens hit the implicit cache (best case; the real hit rate depends on your requests):
| Model and period | Cached input | Uncached input | Output | Monthly total |
|---|---|---|---|---|
| Carbon/Argon (through 2026-11-30) | 40M × $0.10 = $4 | 10M × $2 = $20 | $100 | $124 |
| Carbon/Argon (from 2026-12-01) | 40M × $0.20 = $8 | 10M × $4 = $40 | $200 | $248 |
| Neon | 40M × $0.15 = $6 | 10M × $3 = $30 | $150 | $186 |
| Hydrogen | 40M × $0.075 = $3 | 10M × $1.50 = $15 | $75 | $93 |
⚠️ Three things the math shows: First, from 12-01 Carbon's unit price is 2.67 times Hydrogen's (only 1.33 times through the end of November), while Neon is always 2 times Hydrogen. Second, from 2027-01-01 Gemini 3.8 Flash's unit price ($1.50/$7.50) matches Hydrogen's, so the gap between them before 2026-12-31 comes only from 3.8 Flash's introductory rate. Third, budget with the prices from 12-01; Carbon's and Argon's $2/$10 lasts only through 2026-11-30.
Switching from Argon or 3.8 Flash: What to Watch
- Change the model ID:
gemini-4-carbon,gemini-4-neon, orgemini-4-hydrogen - Set the thinking level again: Neon and Hydrogen default to
high, while Argon defaults tomedium; if you switch from Argon without setting it explicitly, the default level changes and so do output tokens (thinking included). Set it explicitly ingeneration_configand measure tokens again - Decide whether to use
xhigh: Neon and Hydrogen add a fourth level,xhigh; try it only whenhighis not enough - Remove explicit caching: all four Gemini 4 models support implicit caching only, so code that creates or references
cachedContentsmust be rewritten - Only Hydrogen has a free tier: to test Gemini 4 for free, Hydrogen is the only option; note that free-tier content is used to improve Google's products
- Carbon bills double in December: November's bill is not a baseline for December onward
- Hydrogen has official guidance for moving from 3.8 Flash: the changelog's Hydrogen entry links to migration guidance from Gemini 3.8 Flash; read it before switching
Other changes in the same period:
- The docs URL changed: the Gemini API docs moved to aistudio.google.com/docs, and old ai.google.dev/gemini-api/docs URLs redirect there; update bookmarks and links in internal docs
- Gemini 2.5 Pro and 2.5 Flash will be routed automatically: from 2026-12-01, requests to
gemini-2.5-prowill route togemini-3.8-flashandgemini-2.5-flashtogemini-3.5-flash-lite, billed at the new models' prices
CloudInsight Enterprise Gemini API Procurement | Local Invoices, No Foreign Credit Card Needed
The four Gemini 4 models each price differently, and Carbon and Argon change prices at month's end. What enterprises get stuck on most is often not the technology but payment, invoicing, and budget control across billing months.
Contact CloudInsight for a Gemini API Enterprise Quote
We offer one-stop procurement for Claude, GPT, and Gemini, Taiwan uniform invoices, and technical support in Chinese.
How to Choose: Carbon, Neon, Hydrogen, and Argon
| Workload | Recommendation | Why |
|---|---|---|
| Hardest reasoning, sustained agentic coding, cybersecurity defense | Gemini 4 Carbon | Official "strongest reasoning" positioning; $4/$20 from 12-01 |
| Complex work where speed matters (end-to-end task completion) | Gemini 4 Neon | Officially frontier-level with fast completion; $3/$15, cheaper than Carbon from December |
| Gemini 4-level reasoning with fewer tokens than Carbon | Gemini 4 Argon | Officially spends fewer tokens per turn than Carbon; same prices as Carbon |
| Everyday coding, agentic work, new projects | Gemini 4 Hydrogen | Official recommendation for new projects; $1.50/$7.50 |
| Development, testing, proofs of concept | Gemini 4 Hydrogen | The only Gemini 4 model with a free tier |
| High-volume classification, translation, simple data processing | Gemini 3.5 Flash-Lite | The other official choice for new projects; $0.30/$2.50 |
In practice you can tier the work: send everyday traffic to Hydrogen, escalate what Hydrogen can't handle to Neon, and give only the hardest work to Carbon; start with the default thinking level and adjust as needed.
FAQ: Gemini 4 Carbon, Neon, and Hydrogen
How much do Gemini 4 Carbon, Neon, and Hydrogen cost?
Per million input/output tokens: Carbon $2/$10 through 2026-11-30 and $4/$20 from 2026-12-01; Neon $3/$15; Hydrogen $1.50/$7.50. Output prices include thinking tokens; Batch and Flex are half price.
Which one has a free tier?
Only Hydrogen. Carbon, Neon, and Argon are not available on the free tier. Content on Hydrogen's free tier is used to improve Google's products; paid-tier content is not.
Has Gemini 4 Argon been retired?
No. Argon remains available with no shutdown date, and its prices are unchanged. What changed is its positioning: "strongest reasoning" now belongs to Carbon, and Argon is the frontier reasoning model with faster responses and fewer tokens per turn.
Carbon and Argon cost the same. What's the difference?
Their official positioning differs: Carbon is the strongest reasoner; on the same kinds of tasks, Argon responds faster and spends fewer tokens per turn than Carbon. At the same unit price, fewer tokens means a smaller bill, but test with your own tasks.
Which model should a new project use?
Google recommends Gemini 4 Hydrogen or Gemini 3.5 Flash-Lite for new projects. When you need stronger reasoning, evaluate Neon or Carbon.
What is the xhigh thinking level?
A fourth thinking level introduced with Hydrogen, above low, medium, and high; Neon supports it too. Both models default to high. Output prices include thinking tokens, so higher levels usually cost more.
Why did the Gemini API docs URL change?
The official docs moved to aistudio.google.com/docs, and old ai.google.dev/gemini-api/docs URLs redirect automatically. Treat the new URL as the reference.
Conclusion: Default to Hydrogen, Save Carbon for the Hardest Work
Three Core Recommendations
- Start new projects on Hydrogen: it is Google's recommendation, has a free tier, and its unit price is under 40% of Carbon's from December
- Budget with the prices from 12-01: Carbon's and Argon's $2/$10 lasts only through 2026-11-30, then becomes $4/$20
- Tier your usage and set the thinking level explicitly: route by difficulty across Hydrogen, Neon, and Carbon; Neon and Hydrogen default to
high, so measure tokens again when switching models
Ready to Adopt Gemini 4?
Contact the CloudInsight sales team and we'll help plan your Carbon, Neon, and Hydrogen usage mix and handle payment and invoicing.
Join our LINE Official Account for instant model-selection consultation.
Further Reading
- Gemini 4 Argon Guide
- Gemini API Pricing Guide
- What Is Gemini API? Complete Guide
- Gemini 2.5 Pro and Flash Auto-Route From December 1: Which Model You Get, Pricing, and the Code to Change
- GPT-6.2 Astra and Sol Guide
- Claude Fable 5.2 Guide
- LLM API Cost Optimization
References
- Gemini API models (checked 2026-11-19)
- Gemini Developer API pricing (checked 2026-11-19)
- Gemini API release notes (changelog) (checked 2026-11-19)
Need Professional Cloud Advice?
Whether you're evaluating cloud platforms, optimizing existing architecture, or looking for cost-saving solutions, we can help
Book Free ConsultationRelated Articles
Gemini 4 Argon Guide: API Pricing (Intro Rate Through Nov 30), Specs, and Migrating from Gemini 3
Google made Gemini 4 Argon generally available (GA) to developers on 2026-10-13; since 11-10 the strongest-reasoning-model slot belongs to Gemini 4 Carbon, and Argon remains available. The introductory price of $2/$10 only runs through 2026-11-30; from 12-01 it moves to $4/$20, and there is no free tier. This guide covers the official specs (1M context, output up to 1M, three thinking levels), both price stages for every consumption option, cost math, how it splits work with Gemini 3.8 Flash, and the settings to change when moving from the Gemini 3 family.
AI APIGemini 2.5 Pro and Flash Auto-Route From December 1: Which Model You Get, Pricing, and the Code to Change (2026)
On 2026-11-02 Google announced that starting 2026-12-01, every API request to gemini-2.5-pro will automatically route to gemini-3.8-flash and every request to gemini-2.5-flash to gemini-3.5-flash-lite, billed at the new models' rates. This is not a shutdown, and requests won't fail because of the model name, but function calling validation, thinking settings, sampling parameters, and prefilled turns all have breaking changes. This guide covers the routing map, prices before and after, cost math, and the code to change before go-live.
AI APIClaude Fable 5.2 Guide (2026): API Pricing, Differences from Fable 5.1, Mythos 5.2, and Migration
Anthropic released Claude Fable 5.2 on 2026-11-09 at $8/$40 per million input/output tokens, 20% less than Fable 5.1's $10/$50, with cache and batch prices lowered by the same proportion and the full 1M-token context at standard pricing. This guide covers official specs and prices, Mythos 5.2 (same specs and price, verified organizations only), cost math, and the changes that return 400 errors when you move over from Fable 5.1 or Opus 5.5.