Gemini 4 Hydrogen vs Gemini 4 Pro: One of These Models Exists, and Google Is Retiring the Other One Into a Flash Model

The most informative fact about the Pro tier is a routing change buried in Google's deprecation page. Effective 1 December 2026, every API request to gemini-2.5-pro is served by gemini-3.8-flash and billed at Gemini 3.8 Flash rates. The last general-purpose model Google ever shipped under the Pro label will, in three weeks, be answered by a throughput model that scores 31 on the Artificial Analysis Intelligence Index. Meanwhile Gemini 4 Hydrogen — generally available since 12 November 2026 at $1.50 and $7.50 per million tokens — scores 54 at xhigh and, at its own default, 51. Gemini 4 Pro does not exist and never has: there is no such identifier on the rate card, in the model list, or in the documentation. So the comparison is not between two models. It is between a model you can call this afternoon and a label that Google is in the process of retiring.

Every Pro-labelled general model is retired, retiring, or unscoreable

Google's deprecation table is short and it makes the pattern plain. Four identifiers have carried the Pro name for general text work in the past year, and none of them is a live, scored, recommended model today.

• gemini-3-pro-preview — released 18 November 2025, shut down 9 March 2026, replaced by gemini-3.1-pro-preview. Called anyway by a great deal of code that never read the deprecation page.

• gemini-3.1-pro-preview — released 19 February 2026, no shutdown date announced, still priced at $2.00 / $12.00 per million tokens up to 200K and $4.00 / $18.00 above it. It is a preview with no expiry, which is a strange thing to be for nine months, and Artificial Analysis publishes no Intelligence Index figure for it at all — only 113 output tokens per second and 15.28 seconds to first answer token.

• gemini-2.5-pro — deprecated, no longer selectable in Google AI Studio, API access limited to accounts that have actively used it. On 1 December 2026 it stops being a Pro model in any practical sense, because the requests are answered by Gemini 3.8 Flash.

• Nano Banana Pro — gemini-3-pro-image. Real, current, and an image model. It is the only live Pro identifier Google has, and it is not what anyone typing "Gemini 4 Pro" is looking for.

Read those four lines as a sequence and the Pro tier has a direction: preview, deprecation, re-routing to a Flash model. That is not a roadmap. It is a wind-down.

What replaced it, in Google's own words

The Gemini 4 generation shipped three models and gave them element names instead of position names. Google's own selection guidance, from the Hydrogen migration documentation, is unusually direct about the split.

• Gemini 4 Carbon — "our strongest frontier model", spending more thinking tokens per turn to test hypotheses and verify edge cases on hard architecture, debugging, science and security audits. $2.00 / $10.00 through 30 November, $4.00 / $20.00 from 1 December. No free tier, no Contributor tier. This is the model that occupies the slot a Pro model used to occupy.

• Gemini 4 Argon — the efficient frontier model, for work that needs depth but not the ceiling. Same $2.00 / $10.00 introductory card and the same 1 December doubling. This occupies the slot a Pro model used to occupy on a good day, which is precisely the ambiguity that made the label worth dropping.

• Gemini 4 Hydrogen — "recommended upgrade from Gemini 3.8 Flash and earlier", $1.50 / $7.50 flat with no dated step, four thinking levels, a rate-limited free tier, and now the default model behind the Antigravity agent in Gemini Managed Agents.

The price geometry is the tell. When a family has a ceiling at $4 / $20 and a workhorse at $1.50 / $7.50, the middle of the ladder at $2 / $10 is occupied by two models that become identical in price on the same date. A tier name that describes the middle rung has nothing left to describe.

The comparison you can actually run

Since there is no Gemini 4 Pro, the only Pro-labelled model a reader can put a request against is gemini-3.1-pro-preview, and the only honest comparison is on the handful of dimensions where both it and Gemini 4 Hydrogen publish a number.

• Price per million tokens — Hydrogen $1.50 / $7.50 flat across the full window, cache reads $0.075; Gemini 3.1 Pro Preview $2.00 / $12.00 to 200K and $4.00 / $18.00 above, cache reads $0.20 with a 90% discount. Output costs 60% more at the low band and 140% more past 200K.

• Output speed — Hydrogen 190 tokens per second at xhigh and 179 at its default of high; Gemini 3.1 Pro Preview 113. Hydrogen is 58% to 68% faster on generation.

• Time to first answer token — Gemini 3.1 Pro Preview 15.28 seconds; Hydrogen 37.8 at xhigh and 16.7 at high. This is the one line where the old Pro model wins outright, and it wins by 1.4 seconds even against Hydrogen's default. The gap is 22 seconds if you leave Hydrogen at the setting its launch coverage quoted.

• Context and modality — both carry a 1M-token window; Gemini 3.1 Pro Preview accepts text, image, speech and video and outputs text, and Hydrogen takes text, image, video, audio and PDF. Not a differentiator in 2026.

• Independent score — Hydrogen 54 at xhigh, 51 at high, 41 at medium; Gemini 3.1 Pro Preview none published. That blank is the whole problem with keeping a Pro deployment running on faith: the tier's last live model has a latency measurement and no quality measurement, so nobody can say whether the 60% output premium buys anything.

What the four priced lines say together is that the Pro-labelled option is now the slower, dearer, unmeasured one on every axis except the first second and a half of latency. That is a narrow enough advantage to name precisely: it is the advantage of a model that starts answering immediately, which matters for an autocomplete and much less for anything that runs unattended.

What "Pro" was buying, and why nothing bought it back

This part is our reading rather than Google's, and it is worth labelling as such. Tier names describe a position: Flash is throughput, Pro is general capability, Ultra is the ceiling. Position names work when the ladder is monotonic — more money, more capability — because a buyer can pick a rung and trust the ordering.

The Gemini 4 family broke the monotonicity. Gemini 4 Hydrogen, the cheapest member, gives up nine index points to Gemini 4 Carbon and none to Gemini 4 Argon, yet it costs less to finish an index task than either ($1.87 against $7.76 and $4.97 in Artificial Analysis's runs) and runs at two and a half times their throughput. Once a cheap model is the best economics in the family and a dear model is the best capability, the middle rung stops being a decision and starts being a compromise. Element names are what you use when the ladder is not a ladder.

There is a second reason, less flattering. A preview identifier with no shutdown date and no independent score is a way of not committing. Google has kept gemini-3.1-pro-preview priced and callable for nine months without either promoting it or retiring it, while shipping three Gemini 4 models around it. Whatever the internal reason, the effect on a customer is that the Pro string is the one part of the stack nobody has promised anything about.

If you are holding a Pro string today

Four situations, four different actions, and only one of them is urgent.

• gemini-2.5-pro in production — this is the urgent one. Test against Gemini 3.8 Flash before 1 December, because that is what will answer your requests afterwards whether or not you changed anything, at $0.75 / $3.75 instead of the Pro rate. A silent model swap is cheaper and faster and also a different model; if your prompts were tuned to 2.5 Pro's habits, the tuning is the thing that breaks, not the endpoint.

• gemini-3.1-pro-preview in production — re-measure rather than migrate. It has held the latency edge for nine months and it is the only model in this comparison that starts answering in under sixteen seconds without being asked nicely. But it is unscored on the current index, charges $12.00 per million output tokens against Hydrogen's $7.50, and has no announced future. Run your own eval against Hydrogen at high and against Gemini 4 Argon at medium before deciding that 15 seconds is worth an unmeasured model.

• A Pro string in a document rather than a config — replace it with the decision it stood for. Planning and architecture turns go to gemini-4-carbon; execution, tool calls and parallel subagents go to gemini-4-hydrogen; anything already deployed on Argon stays on gemini-4-argon until the eval says otherwise. Google documents the Carbon-plans-Hydrogen-executes pattern itself, including switching models between turns of one session with previous_interaction_id.

• Nano Banana Pro — nothing to do. It is an image model, it is current, and the naming collision is the only reason it appears in this article.

That third action is the one that benefits from a routing layer rather than a model choice, because "Carbon plans, Hydrogen executes" is a composition, not a purchase. OrcaRouter carries Gemini 4 Carbon and Gemini 4 Argon at $2.00 / $10.00 with provider list price passed through at 0% markup, so the 1 December doubling of both cards appears here the same day it appears on Google's, and it carries gemini-3.1-pro-preview at its two-band rate for as long as that preview stays callable — which means a migration off a Pro string needs no second contract. The routing DSL is what lets a planner-and-reviewer split stay in one call instead of in application code, and model fusion lets Carbon and Argon answer the same hard question as a panel rather than as a choice. Gemini 4 Hydrogen is not in our catalogue; it comes from Google's own API and AI Studio, and calling it direct alongside the routed models is the honest arrangement.

The test that would settle it

If a Gemini 4 Pro is ever real, three mechanical things will be true within a day of each other: a gemini-4-pro identifier appears under the Gemini 4 heading in Google's model list, the rate card gains a row with a published figure, and a model card appears carrying a context length and a knowledge cutoff. Any one alone means nothing — an identifier with no price is a gated programme, and a price with no card is a quota reservation. This is checkable in a browser in under a minute, which is more than can be said for any benchmark rumour.

Until then the useful date is 1 December 2026, and it is useful twice over. Carbon and Argon double on it. On the same day, the last consumer Pro model stops being served by a Pro model. If both happen as published, the family Google is selling in 2027 has a $4 / $20 ceiling, a $1.50 / $7.50 workhorse, and no middle — and the correct response to a question about Gemini 4 Pro is not "wait for it" but "you are asking about a tier Google has already replaced with a name."