Key pricing & routing takeaways
Official Gemini API pricing & specs table
Published USD rates per 1M tokens from Google Gemini API Pricing (November 19, 2026). All three models count thinking tokens and grounding tool declarations as billed tokens.
| Parameter / Tier | Gemini 4 Hydrogen ( gemini-4-hydrogen) | Gemini 4 Neon ( gemini-4-neon) | Gemini 4 Carbon ( gemini-4-carbon) |
|---|---|---|---|
| Free API Tier | Available (Free of charge; inputs/outputs used to improve Google products) | Not available on Free tier (available on Paid API and Google AI Pro/Ultra in AI Studio) | Not available on Free tier |
| Standard Input (Paid) | $1.50 / 1M tokens | $3.00 / 1M tokens | $2.00 through Nov 30, 2026 $4.00 starting Dec 1, 2026 |
| Standard Output (incl. thinking) | $7.50 / 1M tokens | $15.00 / 1M tokens | $10.00 through Nov 30, 2026 $20.00 starting Dec 1, 2026 |
| Implicit Context Caching | $0.075 / 1M tokens (Implicit only; explicit unsupported) | $0.15 / 1M tokens (Implicit only; explicit unsupported) | $0.10 through Nov 30, 2026 $0.20 starting Dec 1, 2026 |
| Batch / Flex Tiers | $0.75 in / $3.75 out / $0.0375 cache | $1.50 in / $7.50 out / $0.075 cache | $1.00 / $5.00 / $0.05 (to Nov 30) $2.00 / $10.00 / $0.10 (from Dec 1) |
| Priority Inference | $2.70 in / $13.50 out / $0.135 cache | $5.40 in / $27.00 out / $0.27 cache | $3.60 / $18.00 / $0.18 (to Nov 30) $7.20 / $36.00 / $0.36 (from Dec 1) |
| Context & Max Output | 1,048,576 unified context; 65,536 sync output / up to 1,048,576 long decode | 1,048,576 unified context; 65,536 sync output / up to 1,048,576 long decode | 1,048,576 unified context; 65,536 sync output / up to 1,048,576 long decode |
| Thinking Controls & Modalities | low, medium, high (default), xhigh; Text, Image, Video, Audio, PDF → Text | low, medium, high (default), xhigh; Text, Image, Video, Audio, PDF → Text | Configurable thinking levels; Text, Image, Video, Audio, PDF → Text |
| Search & Maps Grounding (Paid) | 5,000 free/mo shared across Gemini 3+ (1,500/day cap), then $14 / 1,000 queries | 5,000 free/mo shared across Gemini 3+ (1,500/day cap), then $14 / 1,000 queries | 5,000 free/mo shared across Gemini 3+ (1,500/day cap), then $14 / 1,000 queries |
| Paid Data Training Policy | No (Paid tier data is not used to improve Google products) | No (Paid tier data is not used to improve Google products) | No (Paid tier data is not used to improve Google products) |
Where Gemini 4 Neon fits — and what to watch before Dec 1
1. Capabilities on gemini-4-neon: Per the Gemini 4 Neon model specification, Neon supports function calling, structured outputs, code execution, computer use, file search, URL context, Google Search/Maps grounding, and agentic video understanding (static default). It does not support image generation, audio generation, or the Live API.
2. Thinking tokens dominate real bills: Every Gemini 4 tier bills internal thinking tokens at the output rate ($15/1M on Neon, $7.50/1M on Hydrogen, $10→$20/1M on Carbon). Setting low or medium thinking on routine requests instead of leaving the default high setting often trims latency and spend more than switching base models.
3. The Nov 19–30 Carbon pricing overlap: Because Google discounted Gemini 4 Carbon and Argon through November 30, 2026 ($2.00 in / $10.00 out), Carbon is temporarily cheaper per token than Neon ($3.00 / $15.00). Teams planning December production budgets should benchmark against Carbon’s $4.00 / $20.00 standard rate rather than the introductory rate.
4. Implicit caching only: None of Hydrogen, Neon, or Carbon currently supports explicit context caches; repeated prefixes rely on automatic implicit caching (a 95% discount relative to each model’s standard input rate: $0.075, $0.15, and $0.10→$0.20 per 1M cached tokens).
Decision checklist: Hydrogen vs Neon vs Carbon
Pick Gemini 4 Hydrogen ($1.50 / $7.50) when: you want free-tier experimentation in prototyping, run high-volume tool-calling agents, or want half the token cost of Neon on everyday coding and document QA.
Pick Gemini 4 Neon ($3.00 / $15.00) when: Hydrogen hits reasoning ceilings on multi-file coding, knowledge synthesis, or science workflows, and you want faster end-to-end completion than Carbon at 25% below Carbon’s post-December 1 standard price ($4.00 / $20.00).
Pick Gemini 4 Carbon ($2/$10 to Nov 30; $4/$20 from Dec 1) when: you are solving hard system-architecture, security, or deep scientific verification tasks where spending extra thinking tokens for edge-case verification justifies top-tier output rates.
Related comparisons and official documentation: Gemini 4 Hydrogen vs Claude Sonnet 5.5 · How to access Gemini 4 Carbon · Google Gemini overview · What’s New in Gemini 4 Neon (Google AI Studio) · Choosing a Gemini Model · Gemini 4 Neon Model Card.
Frequently asked questions
What is Gemini 4 Neon API pricing per 1M tokens?
Gemini 4 Neon costs $3.00 per 1M input tokens, $15.00 per 1M output tokens (including thinking tokens), and $0.15 per 1M cached input tokens on the Standard Paid tier. Batch and Flex tiers cost $1.50 / $7.50 / $0.075, and Priority inference costs $5.40 / $27.00 / $0.27. There is no Free API tier.
Why is Gemini 4 Carbon cheaper than Gemini 4 Neon before December 1, 2026?
Gemini 4 Carbon (and Argon) carry introductory pricing of $2.00 input / $10.00 output per 1M tokens through November 30, 2026. Starting December 1, 2026, Carbon standard rates double to $4.00 input / $20.00 output per 1M tokens.
Which Gemini 4 model includes a Free API tier?
Only Gemini 4 Hydrogen offers a Free API tier (where prompts and responses may be used to improve Google products). Gemini 4 Neon and Carbon are Paid-tier models on the API, though Neon is available on Google AI Pro and Ultra plans in Google AI Studio.
What context window and caching rules apply across Gemini 4 Neon, Hydrogen, and Carbon?
All three models offer a unified 1,048,576-token context window (65,536 synchronous max output, or up to 1,048,576 tokens with long decode) and support implicit context caching only (explicit caching is unsupported).
Pricing and model specs reflect published Google AI Studio documentation as of November 19, 2026; verify live rate limits, regional endpoints (Vertex AI non-global endpoints differ), and thinking-token usage before production rollout.
Real-time interactive world model API: DeepMind Genie 4 API Pricing & Limits Review ($17.50/1M, ~$6.05/min) · Local 32GB open-weight coding model: Inkling 2 Mini vs Qwen3.8-27B & Gemma 4 31B Benchmarks