Google made Gemini 4 Neon (model ID gemini-4-neon) generally available on November 18, 2026. It is the fourth Gemini 4 model after Argon (October 13), Carbon (November 10) and Hydrogen (November 12), and Google places it between Hydrogen and Carbon.
The short version: Neon costs $3.00 per 1M input tokens and $15.00 per 1M output tokens, with no promotional price. It is not on the Gemini API free tier; in Google AI Studio it is included with AI Pro and AI Ultra. The biggest decision point: when Carbon moves to $4.00 / $20.00 on December 1, Neon becomes 25% cheaper than Carbon at the same token count. Everything below comes from the official Gemini API changelog, pricing page, Choosing a model and the "What's new in Gemini 4 Neon" guide. Checked: November 22, 2026.
Gemini 4 Neon at a glance
Item | Details |
|---|---|
Model ID |
|
Released | GA on November 18, 2026 |
Price (Standard) | $3.00 input / $15.00 output per 1M tokens (thinking billed as output), flat across 1M context, no promo |
Caching | Implicit only; cached input $0.15 (95% off). Explicit caching not supported |
Batch / Flex | $1.50 / $7.50 |
Priority | $5.40 / $27.00 |
Free tier | None on the API. AI Studio: AI Pro and AI Ultra |
Context | 1,048,576 tokens shared by input, thinking and output |
Output | 65,536 synchronous; up to ~1M with long decode (background execution) |
Thinking levels | low / medium / high (default) / xhigh; minimal returns an error |
Not supported | Image generation, audio generation, Live API |
Carbon vs Neon vs Hydrogen
Model | Speed (official) | Thinking (default) | Price (in / out) | AI Studio plans |
|---|---|---|---|---|
Gemini 4 Carbon | Deliberate | low–high (medium) | $2.00 / $10.00 through Nov 30, then $4.00 / $20.00 | AI Ultra |
Gemini 4 Neon | Moderate | low–xhigh (high) | $3.00 / $15.00 | AI Pro, AI Ultra |
Gemini 4 Hydrogen | Fast | low–xhigh (high) | $1.50 / $7.50 | All plans incl. free |
Gemini 3.5 Flash-Lite | Fastest | minimal–high (minimal) | $0.30 / $2.50 | All plans incl. free |
Google recommends Neon for multi-file refactoring, repository-wide debugging and analytical work where a developer or coding agent is waiting on the answer: deeper reasoning than Hydrogen, and faster with fewer output tokens than Carbon. Carbon is for multi-hour background runs, deep research, proofs and security audits, or as the step up when Neon at high or xhigh gets stuck. Hydrogen stays the default for everyday coding, document pipelines and parallel subagents.
Neon or Carbon: the answer flips on December 1
Example: 100 code-investigation tasks a month at 200K input and 30K output (including thinking) tokens each, Standard pricing, no caching.
Model | 20M input | 3M output | Monthly |
|---|---|---|---|
Carbon (November promo) | $40 | $30 | $70 |
Neon | $60 | $45 | $105 |
Carbon (from Dec 1) | $80 | $60 | $140 |
Hydrogen | $30 | $22.50 | $52.50 |
- In November, Carbon is cheaper unless Neon uses two-thirds or less of Carbon's tokens.
- From December 1, Neon is cheaper: 75% of Carbon at equal tokens, and the gap grows if Neon really needs fewer output tokens.
- Neon is exactly 2x Hydrogen, so keep Hydrogen wherever its high level already meets your bar.
If half of the input hits the implicit cache, Neon's input cost drops from $60 to about $31.50. Non-urgent jobs on Batch or Flex cost $1.50 / $7.50, the same as Hydrogen Standard.
Where you can use it
- Gemini API: paid-tier projects only, on both generateContent and the Interactions API.
- Google AI Studio: per Google's AI plans page, Neon is on AI Pro and AI Ultra; Carbon needs AI Ultra; Hydrogen is on every plan.
- Managed Agents: Antigravity still defaults to Hydrogen; set Neon in
agent_config.
Migration: Neon replaces 3.1 Pro Preview
Google's upgrade table now maps gemini-3.1-pro-preview to gemini-4-neon (or Carbon for the hardest problems). From Gemini 3 and earlier: change the model ID, raise low max_output_tokens caps to 65,536, expect high as the default thinking level (minimal errors), drop explicit caching and temperature/top_p/top_k, and declare tools natively. From Carbon, Argon or Hydrogen, only the model ID changes, but note Carbon and Argon default to medium while Neon defaults to high.
FAQ
How much does Gemini 4 Neon cost?
On the Gemini API paid tier (Standard), Neon costs $3.00 per 1M input tokens and $15.00 per 1M output tokens, including thinking tokens, flat across the full 1M context. Implicit cache hits cost $0.15 per 1M input tokens, and Batch and Flex cost $1.50 input and $7.50 output.
Can I use Gemini 4 Neon for free?
Not on the Gemini API free tier. In Google AI Studio, Neon is available on the AI Pro and AI Ultra plans. Hydrogen is the only Gemini 4 model on the free tier.
Is Gemini 4 Neon cheaper than Carbon?
Through November 30, 2026, Carbon's promotional price of $2.00 / $10.00 is lower per token. From December 1, Carbon costs $4.00 / $20.00, so at the same token count Neon is 25% cheaper. Google also says Neon finishes with fewer output tokens than Carbon.
What should I migrate gemini-3.1-pro-preview to?
Google's upgrade table recommends gemini-4-neon, or gemini-4-carbon for your hardest problems.