Gemini 4 Neon
Language Available Full comparison ↗ google-deepmind/gemini-4-neon · by Google DeepMind
· decoder-only-transformer
Pricing — 1 offering(s)
About this offering: Gemini 4 Neon, Google's Gemini 4 model for complex software engineering, knowledge work and science, generally available on the Gemini API since 2026-11-18 (paid tier only; not on the Free tier). It sits between Gemini 4 Hydrogen and Gemini 4 Carbon. One price at every prompt length up to its 1M-token context, with no >200K long-context band. Batch is half price ($1.50 / $7.50, cached input $0.075). Flex is also 50% of Standard, and Priority is 1.8x Standard ($5.40 / $27, cached input $0.27) (Google pricing page, checked 2026-11-19).
Input tokens
- $3.00 / 1M tokens (input) Current 2026-11-18 → present
Cached input tokens
- $0.15 / 1M tokens (cached input)
Current
2026-11-18 → present
Implicit caching only (explicit caching is not supported); 95% off input.
Output tokens
- $15.00 / 1M tokens (output)
Current
2026-11-18 → present
Includes thinking tokens.
Batch
Batch input tokens
- $1.50 / 1M tokens (input) Current 2026-11-18 → present
Batch output tokens
- $7.50 / 1M tokens (output) Current 2026-11-18 → present
Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.
See how Gemini 4 Neon fits into a cost-aware routing setup
See how →Capability profile
How Gemini 4 Neon rates across core capability dimensions, with the task-level evidence behind each rating.
Four tunable thinking levels (low, medium, high by default, and xhigh) for deeper reasoning on hard multi-step problems, per Google's What's new page.
Google's What's new page: built for complex software engineering, completing repository-level debugging and multi-file refactoring in fewer steps. Google's launch post cites Artificial Analysis's Coding Agent Index at 58.6 (#1 overall), as reported by Google. No independent coding benchmark recorded yet.
Function calling, code execution, computer use, file search, URL context and Google Search/Maps grounding are listed as supported on Google's model page. Google says it reduces redundant file reads and handles context compaction reliably across long sessions with many tool calls.
1,048,576-token unified input and output window; 65,536 output tokens by default, up to about 1M with long decode.
Google positions it for fast end-to-end task completion with leaner token use, between the faster Gemini 4 Hydrogen and the deeper Gemini 4 Carbon.
$3/$15 per 1M tokens at every prompt length: twice Gemini 4 Hydrogen ($1.50/$7.50), and above Gemini 4 Carbon and Argon until 2026-11-30 ($2/$10), below them from 2026-12-01 ($4/$20).
Ratings are estimated — limited independent data is available for this model.
Operator guidance
Google places Neon between Hydrogen and Carbon: keep Hydrogen as the default for everyday coding and agentic work, use Neon for complex software engineering, knowledge work and science workflows, and step up to Carbon for the hardest architecture, debugging, science and security problems. Until 2026-11-30 Carbon and Argon ($2/$10) cost less per token than Neon ($3/$15); from 2026-12-01 they cost $4/$20, so Neon becomes the cheaper of the three.
Use cases
- Complex software engineering: repository-level debugging and multi-file refactoring
- Knowledge work and science: synthesising long documents, spreadsheets and charts into structured deliverables
- High-frequency complex tasks where Gemini 4 Carbon is slower or more than you need
Limitations
- Implicit caching only; explicit context caching is not supported
- Paid tier only; not available on the Gemini API Free tier
- Capability ratings are qualitative, from Google's own materials, not a cited independent benchmark run