Model comparison
Gemini 4 Neon vs Argon
Gemini 4 Neon is designed for fast iteration on complex work. Gemini 4 Argon is a useful baseline for advanced coding and deeper research, especially if your workflow already uses it.
Which model should you use?
Start with Neon for repeated patch reviews and short feedback loops. Compare Argon on repository-wide changes or long research tasks where the quality of the final result matters more than a fast first reply.
Example: For a failing integration test, ask Neon to pinpoint the regression and propose a patch. Use Argon for a wider investigation that traces upstream assumptions across the repository and flags dependencies needing review.
Standard API pricing
USD per million tokens. Output charges include thinking tokens. Prices checked .
| Model | Uncached input | Output | Rate period |
|---|---|---|---|
| Neon | $3 | $15 | Standard rate at time of check |
| Argon | $2 | $10 | Through Nov 30, 2026 |
| Argon | $4 | $20 | From Dec 1, 2026 |
For 1 million uncached input tokens and 100,000 output tokens, the estimate is $4.50 for Neon, $3 for Argon during the promotion, and $6 at Argon’s scheduled standard rate. Cache hits and tool fees change the total. Calculate Neon API cost
How to compare them on your work
Give both models the same failing build log, relevant source files, and test suite. Ask for a patch and a short explanation. Compare test pass rate, missed regressions, elapsed time, and paid tokens across retries.
Google announced Neon as an adaptation of Carbon. A family relationship does not establish how Neon compares with Argon on your own repository. Use the same tool permissions and a fixed test harness for both.
Sources checked November 19, 2026. Google Gemini Developer API pricing, Google DeepMind model card and Artificial Analysis.