Google brings Gemini 4 Hydrogen to AI Mode and Search, replacing Gemini 3.8 Flash
Google released Gemini 4 Hydrogen, a fast, lower-cost model that replaces Gemini 3.8 Flash and powers what Google calls its biggest Search model improvement yet. Outlets agree on the headline claims but differ on benchmark details and on who gets access in the Gemini app.
1 / 2
The story, neutrally told
Mixed · 2Google has released Gemini 4 Hydrogen, a model positioned as the fast, economical tier that replaces the Flash models. Ars TechnicaLC “Hydrogen’s purpose is clear: It replaces the faster, more economical Flash models most people in Google’s ecosystem use.” Read at Ars Technica ↗ TechCrunchN “Gemini 3.8 Flash, its previous fast model” Read at TechCrunch ↗ Left · 1Ars Technica reports that Hydrogen has replaced Gemini 3.8 Flash across Google's products and that access began immediately, unlike the staged rollouts of the Carbon and Argon models, which began with cybersecurity researchers and were first limited to the most expensive subscription. Ars TechnicaLC “The limits for Hydrogen are much more relaxed, with access beginning immediately. This model has replaced Gemini 3.8 Flash across the company’s products.” Read at Ars Technica ↗ Mixed · 2In Search, Hydrogen is meant to better understand complicated questions, including multipart or highly specific requests and typos, according to Google, which calls it its "biggest model improvement to Search yet". TechCrunchN “Google claims is its “biggest model improvement to Search yet.”” Read at TechCrunch ↗ Ars TechnicaLC “The new model should better handle long, multipart questions” Read at Ars Technica ↗
Mixed · 2Google says AI Overviews and AI Mode will show more visual elements, such as images, structured responses, comparison tables and custom tools, and that Hydrogen improves answers across the languages AI Mode supports. TechCrunchN “more images, structured responses, comparison tables, and custom tools will appear as visual responses” Read at TechCrunch ↗ Ars TechnicaLC “the responses in both AI Overviews and AI Mode will have more visual elements” Read at Ars Technica ↗ Mixed · 2Google says the model's reasoning improves Search's fan-out technique, which splits a prompt into several subqueries, so that it links to more sources; Ars Technica notes that it has heard such claims from Google before. TechCrunchN “improves Search’s fan-out technique, which breaks down a single user’s prompt into multiple subqueries” Read at TechCrunch ↗ Ars TechnicaLC “but we’ve heard that from Google before” Read at Ars Technica ↗ Centre · 1Google says the update is rolling out to all Search users globally. TechCrunchN “Google says the Hydrogen update is rolling out to all Search users globally.” Read at TechCrunch ↗
Mixed · 2Google says Hydrogen is 2.5 times faster than Gemini 3.8 Flash and offers four thinking levels: low, medium, high and a new "xhigh". Ars TechnicaLC “It runs in one of four thinking levels (low, medium, high, and the new xhigh)” Read at Ars Technica ↗ TechCrunchN “Hydrogen completes tasks 2.5 times faster than Gemini 3.8 Flash” Read at TechCrunch ↗ Left · 1Ars Technica adds that, by Google's own analysis, Hydrogen cuts "fabricated claims" by more than 5x and more than triples 3.8 Flash's Terminal-Bench score. Ars TechnicaLC “it reduces “fabricated claims” by more than 5x, according to Google’s analysis”“Hydrogen more than tripling 3.8 Flash’s score in Terminal-Bench” Read at Ars Technica ↗ Mixed · 2On the Artificial Analysis Intelligence Index, both outlets say Hydrogen beats Gemini 4 Argon, but they describe the setting differently: Ars Technica cites 54 versus 50 in xhigh mode, while TechCrunch says it did so at the default high level. Ars TechnicaLC “It managed a score of 54 in the popular Artificial Analysis Intelligence Index with xhigh mode versus 50 for Argon.” Read at Ars Technica ↗ TechCrunchN “at its default high thinking level” Read at TechCrunch ↗
Mixed · 2Google still presents Carbon and Argon as the choices for deep reasoning, long-horizon tasks and security, with Hydrogen aimed at fast agentic work such as writing code, running tests and verifying changes, alone or paired with the larger models. TechCrunchN “Gemini 4 Carbon and Gemini 4 Argon, are best at deep reasoning for hard problems” Read at TechCrunch ↗ Ars TechnicaLC “it still positions Argon and Carbon as the best for long-horizon tasks and security” Read at Ars Technica ↗ Centre · 1TechCrunch reports that, according to Google, Hydrogen ranks No. 5 on Agent Arena's benchmarks for multi-step tasks. TechCrunchN “Hydrogen ranks No. 5 on Agent Arena’s benchmarks, the company pointed out.” Read at TechCrunch ↗ Mixed · 2Availability outside Search is described differently: TechCrunch says Hydrogen is coming to all tiers of AI Studio, the Gemini API and the Gemini app, and to paid Pro and Ultra users of Antigravity, while Ars Technica says it is live in the Gemini app and AI Studio only for AI Pro and Ultra subscribers, with AI Plus and free users still on 3.5 Flash-Lite. TechCrunchN “coming to all tiers of Google’s AI Studio, the Gemini API, the Gemini App” Read at TechCrunch ↗ Ars TechnicaLC “AI Plus and free users are still stuck with the old 3.5 Flash-Lite model” Read at Ars Technica ↗
Left · 1Ars Technica reports API pricing of $1.50 per million input tokens and $7.50 per million output tokens, unchanged from 3.8 Flash, while Carbon and Argon, now $2/$10, will rise to $4/$20 next month. Ars TechnicaLC “$1.50 per million input tokens and $7.50 per million output tokens, which is unchanged from 3.8 Flash”“they will go up to $4/$20 starting next month” Read at Ars Technica ↗
Every sentence links to the reporting it rests on. The pill in front of each says where its sources sit: Left, Centre or Right when one side supplies at least half of them, Mixed when they are evenly split. The number is how many outlets it cites.
Left1 outlet
- Framing
- Ars Technica treats it as a model launch, comparing Hydrogen with Google's other models on rollout, speed, benchmarks, access limits and pricing, and is sceptical of some of Google's claims.
- Emphasis
- Staged-versus-immediate rollout, subscription tiers, API pricing and the coming price rise for Carbon and Argon; coding use.
- Leaves out or plays down
- Gives little detail on the Search improvements beyond a short paragraph; does not mention Agent Arena or the multilingual gains.
- Charged language
- “finally replacing”“stuck with the old 3.5 Flash-Lite model”
- For example
-
“but we’ve heard that from Google before” — Ars Technica
“AI Plus and free users are still stuck with the old 3.5 Flash-Lite model” — Ars Technica
Centre1 outlet
- Framing
- TechCrunch frames it as a Search upgrade that will give better answers to complicated questions, relaying Google's announcement and blog post, mostly in Google's own terms.
- Emphasis
- Search improvements: intent, typos, fan-out, links, visual responses and languages; the idea that a faster, cheaper model is often enough.
- Leaves out or plays down
- Does not give API pricing, the rollout contrast with Carbon and Argon, or the restrictions in the Gemini app; offers no outside scepticism of Google's claims.
- Charged language
- “biggest model improvement to Search yet”
- For example
-
“Google says the Hydrogen update is rolling out to all Search users globally.” — TechCrunch
“Hydrogen is able to more precisely address each part of a user’s question” — TechCrunch
Right0 outlets
No right outlet in our sources has covered this story yet.
What every side reports
- Gemini 4 Hydrogen replaces Gemini 3.8 Flash as Google's fast model.
- Google says it is 2.5 times faster than Gemini 3.8 Flash.
- It has four thinking levels: low, medium, high and xhigh.
- It powers upgrades to Search and AI Mode, with better handling of complex questions and more visual responses.
- Google says it beats Gemini 4 Argon on the Artificial Analysis Intelligence Index.
Where accounts differ
-
Which setting beat Argon on the Artificial Analysis Intelligence Index
- Left
- Ars Technica: 54 in xhigh mode versus 50 for Argon.
- Centre
- TechCrunch: beat Argon at the default high thinking level.
-
Who can use Hydrogen in the Gemini app and AI Studio
- Left
- Ars Technica: live for AI Pro and Ultra subscribers; AI Plus and free users are still on 3.5 Flash-Lite.
- Centre
- TechCrunch: coming to all tiers of AI Studio, the Gemini API and the Gemini app.
Google organisation
Google presents Hydrogen as a major improvement over 3.8 Flash and the biggest Search model upgrade yet, with Carbon and Argon kept for the hardest tasks.
“Google claims is its “biggest model improvement to Search yet.”” — TechCrunch
“Google claims that Hydrogen represents a major improvement versus the previous Flash model.” — Ars Technica
Google AI Mode product
AI Mode is to give richer visual responses, better answers across languages and more helpful links, and the update is rolling out to all Search users globally.
“better answers across the different languages supported by Google’s AI Mode” — TechCrunch
Gemini 4 Hydrogen product
Hydrogen is a fast, cheaper model with four thinking levels, built for agentic workflows and everyday coding, alone or paired with Carbon and Argon.
“Hydrogen is designed for fast execution across a variety of agentic workflows” — TechCrunch
Left1 article
-
Google rolls out Gemini 4 Hydrogen, finally replacing 3.8 Flash
Mixed Detailed launch report that relays Google's performance claims with some scepticism and stresses access limits and pricing.

Centre1 article
-
Google’s latest Search upgrade promises better answers to your most complicated questions
Neutral Product-announcement piece that relays Google's claims about Search improvements without independent testing.

Right0 articles
No coverage yet.
- 12 Nov 19:23 First Ars TechnicaLC Google rolls out Gemini 4 Hydrogen, finally replacing 3.8 Flash
- 12 Nov 19:27 +5m TechCrunchN Google’s latest Search upgrade promises better answers to your most complicated questions
Times are when each article was published, or when we first saw it if the outlet gave no time.
Ask about this story
Loading…