In the Gemini 4 family, Hydrogen is the fast tier and Neon is the mid-tier, and the price differs by exactly 2x: Neon at $3/$15, Hydrogen at $1.50/$7.50. Both share a 1M context, four thinking tiers and the same input/output modalities. So what does twice the price actually buy? The answer: a 7-point AA index gap (61 vs 54), No. 1 finishes in coding and enterprise knowledge work, and a higher bar to entry — Neon has no free tier. Which one to use day to day depends on whether your work is hard enough to justify paying double for 7 points.
| Item | Neon | Hydrogen |
|---|---|---|
| Standard (input / output, per 1M tokens) | $3 / $15 | $1.50 / $7.50 |
| Cached input | $0.15 | $0.075 |
| Batch / Flex | $1.50 / $7.50 | $0.75 / $3.75 |
| Priority | $5.40 / $27.00 | $2.70 / $13.50 |
| Free tier | No | Yes |
Every tier is exactly 2x. Two equivalences are worth remembering: Neon's Batch / Flex price ($1.50/$7.50) equals Hydrogen's standard price; Hydrogen's Priority price ($2.70/$13.50) is still below Neon's standard price. The free tier is the biggest barrier difference: Hydrogen's API has a free tier and free users can use it in the Gemini app; Neon's API is paid-tier only and the Gemini app needs a Pro / Ultra subscription. For zero-budget trials and light calls, Hydrogen is the only option.
On the AA index v5.0.1: Neon at 61 (xhigh) and 60 (default high); Hydrogen at 54 (xhigh) and 51 (default tier). A 7-point gap at the top tier, 9 points at the default tier. That is the second-largest gap inside the family (Carbon vs Hydrogen is 60 vs 54, a 6-point gap). Neon's lead concentrates in coding and enterprise knowledge work: the AA Coding Agent Index at 58.6, No. 1; the Vals Index at 70.9%, No. 2; AutomationBench at 56.3% (Google-reported). Hydrogen's scores on those individual tests were not verified one by one in this round, so no direct comparison there — but a 7–9 point composite gap already says Neon handles complex multi-step work a tier better.
One caution: Neon's 61 needs the xhigh tier, and one xhigh index run generates 283M output tokens — 2.3x the high tier's; Hydrogen's 54 is likewise its xhigh figure. Both top-tier scores are "most expensive tier" scores; the default-tier gap (60 vs 51) is the more honest comparison.
Both offer low / medium / high / xhigh, both default to high. The difference is not the number of tiers but their quality: Neon is adapted from Carbon, while Hydrogen is the family's fast base (succeeding 3.8 Flash). Google's design intent is clear: four tiers let users switch models under the same control scheme — moving from Hydrogen to Neon needs no new tier vocabulary, just twice the price for higher tier scores.
Google positions Hydrogen as "fast everyday execution" and Neon as "fast end-to-end task completion", but the launch materials contain no same-workload timing comparison between the two. The only known figure is Neon's: AA measured 21.5 minutes per index task and 56.51 seconds to first answer token at xhigh — among the slowest in its price tier. Hydrogen's same-measure timing was not verified in this round, so no guess here. The only fair statement: Neon's "fast" is relative to Carbon; relative to Hydrogen it is the more expensive, more capable and slower one — "fast" for high-frequency complex work means finishing jobs cleanly, not delivering the first token sooner.
In the Gemini app, Hydrogen is available to free, Pro and Ultra users ("Everyday help" is 4 Hydrogen); Neon needs a Pro / Ultra subscription ("Deeper analysis"). On the API, Hydrogen has a free tier and Neon does not. In Antigravity both need paid Pro / Ultra, and Neon can also act as an Orchestrator directing Hydrogen subagents — the combination Google designed: Neon judges, Hydrogen executes.
| Task type | Pick | Why |
|---|---|---|
| Everyday coding, interactive programming, quick loops | Hydrogen | Half the price, has a free tier; Neon's 7-point edge does not show on these tasks |
| Zero-budget trials, light calls | Hydrogen | Neon has no free tier |
| Complex multi-step coding agents, enterprise knowledge work | Neon | AA 61 vs 54; Coding Agent Index No. 1; Vals 70.9%, No. 2 |
| Long-horizon agent work, tough debugging | Neon (or Carbon) | Hydrogen's capability ceiling is lower; the hardest jobs are covered in our Neon vs Carbon comparison |
| Non-urgent batch work | Neon Batch | Batch at $1.50/$7.50 equals Hydrogen's standard price, with a tier more capability |
| Free Gemini app users | Hydrogen | Neon needs a Pro / Ultra subscription |
| Antigravity orchestration | Neon + Hydrogen together | Neon as Orchestrator, Hydrogen as subagents — Google's official combination |
Moving from Hydrogen to Neon means changing only the model ID (gemini-4-hydrogen → gemini-4-neon); the request schema is unchanged. Three things to watch: first, the bill doubles — the same call volume costs exactly 2x, so validate the effect on small traffic before switching everything; second, the free tier disappears — calls that used to run free now need a paid tier and a billing account; third, the tier vocabulary is the same but the scores differ — at high, Neon scores 60 and Hydrogen 51, so do not carry old tier habits over unchanged.
Reference sources: this article is an original rewrite produced by this site; facts and figures were checked against Gemini API pricing (read Nov 19, 2026), Artificial Analysis' model page (read Nov 19, 2026) and Google's official blog (Nov 18, 2026)
← Back to the Gemini 4 Neon topic page