Google

Gemini 4 Hydrogen

Models.dev
Intelligence
53.9
Rank 25 of 813 measured models
Output speed
153 t/s
Rank 76 of 476 measured models
Context window
1M
Rank 120 of 1420 measured models
Cost per task
$3.16
Rank 208 of 255 measured models

Gemini 4 Hydrogen (Xhigh) is a text AI model from Google, released on Nov 12, 2026, with a context window of 1M tokens. Gemini 4 Hydrogen (Xhigh) scores 53.9 on the Artificial Analysis Intelligence Index, ranking 25th of 813 measured models. The closest models on this index are GPT-6.1 Sol (54.6), Claude Sonnet 5.5 (Default Fallback) (54.5), and GPT-6.1 Terra (51.3). With an output speed of 153 t/s, it is faster than GPT-6.1 Sol (46.5 t/s) and GPT-6.1 Terra (65.5 t/s), about as fast as Claude Sonnet 5.5 (Default Fallback) (151 t/s). Every benchmark side by side: Gemini 4 Hydrogen (Xhigh) vs Claude Sonnet 5.5 (Default Fallback).

Artificial Analysis Indices

Extra high
Intelligence53.9
Rank 25 of 813 measured models

Reasoning levels

Each point is a reasoning level of this model: more reasoning often raises the score but costs more. Select a point to open that level.
Intelligence Indexmost efficient ↗Cost per task

Capability indexes

Performance in specific industries, measured by Artificial Analysis
Finance & Accounting58.1
Rank 15 of 283 measured models
Strategy & Ops57.7
Rank 24 of 283 measured models
Legal56.7
Rank 15 of 283 measured models
Healthcare & Medical53.3
Rank 14 of 210 measured models
Engineering60.1
Rank 20 of 281 measured models
Economics55.0
Rank 30 of 289 measured models

Standard Benchmarks

FrontierCode 2.039.0%
Rank 11 of 37 measured models
Terminal-Bench-Science49.5%
Rank 11 of 37 measured models
HLE-Diamond58.8%
Rank 9 of 37 measured models
CritPt v2 (Hard Subset)66.0%
Rank 8 of 33 measured models
AutomationBench-AA76.1%
Rank 5 of 37 measured models
GDP.pdf (all-pass)27.6%
Rank 11 of 37 measured models
FrontierMath Tiers 1–3 (Epoch AI)88.1%
Rank 13 of 82 measured models
FrontierMath Tier 4 (Epoch AI)87.8%
Rank 14 of 69 measured models
Additional benchmarks10
Design Arena · 3d65.8%
Design Arena · Asciiart72.7%
Design Arena · Codecategories56.4%
Design Arena · Codetovideo69.5%
Design Arena · Dataviz54.5%
Design Arena · Gamedev52.3%
Design Arena · Mobileapps60.0%
Design Arena · Pdf54.3%
Design Arena · Uicomponent54.0%
Design Arena · Website58.7%

Models at a similar level

The closest models on the Intelligence Index, one configuration per family.

Preferences and analytics

Rybbit: visits, performance, clicks, copies, forms and session replay (masked inputs), with your permission. Refusing does not limit the site.

Privacy

Allow navigation analytics?

With your permission, Rybbit measures visits, performance and interactions (clicks, copies, forms), including session replay, to improve BenchSift. Input values are masked in replays. Refusing does not limit the site.

Your choice is stored for 180 days. Change it at any time using “Privacy preferences” in the footer. Withdrawal reloads the page.

No choice stored: Rybbit is disabled.

Privacy