whatmodel.sh / Gemini 4 Hydrogen

Gemini 4 Hydrogen is a closed-weights language model from Google, released November 2026. Licence: proprietary.

Catalogue identity notes

Gemini API changelog, November 12, 2026: 'Gemini 4 Hydrogen (gemini-4-hydrogen) generally available (GA)', on the free tier and at paid-tier pricing of $1.50 per 1M input tokens and $7.50 per 1M output tokens. Stable model code gemini-4-hydrogen; thinking levels low, medium, high (default) and xhigh. Google presents it as the successor to Gemini 3.8 Flash. Distinct from Gemini 4 Argon and Gemini 4 Carbon. Arena lists it as 'Gemini 4 Hydrogen (High)'.

Model information sources

Release date ↗

79.1±1.7#14 of 743 LLMsfrom 3 benchmark results across 2 sources

Score within the LMArena (text) comparison set only.

$ how it compares

The LLMs ranked either side of Gemini 4 Hydrogen, out of 897 with a score, then more from Google.

#modelscorewhereRAMreleased
12GPT-5.5 Pro OpenAI80±1api $67.50/M2026-04
13GPT-6 Sol OpenAI80±1api $4.00/M2026-09
14Gemini 4 Hydrogen Google79±2api $3.00/M2026-11
15GPT-5.6 Sol OpenAI79±0api $8.00/M2026-07
16Claude Mythos Preview (limited preview) Anthropic79±1limited preview2026-04
more from Google
3Gemini 4 Carbon Google84±1api $4.00/M2026-10
9Gemini 4 Argon Google81±1api $4.00/M2026-09
35Gemini 3.8 Flash Google75±1api $1.50/M2026-09

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▄▅▅▃▃▂▃ #7 of 319 * · 21% of its score · 1 test

Beats GPT-6 Astra, GPT-6.1 Terra, Claude Fable 5.1 here, though they rank higher overall. Trails Kimi K3.1 Pro here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Terminal-Bench 4.156.4% ✎▂▂▂··▅·▂▅█▃▆ beats 18/20very hard · avg 1.0%▇▇▇▇▇▇▇▇▇ 21%sharp · at its level · noisy
coding▁▁▁▃▆▇█▆▆▃▃▁ #9 of 474 * · 50% of its score · 1 test

Beats GPT-6.1 Terra, Gemini 4 Argon, Claude Opus 5 here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
Arena WebDev (Code Arena)1704.2 ✎▁▂▄▃▇█▅▇▇▅▂▁ beats 126/134avg 996▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 50%sharp
chat & writing▁▁▁▃▅▄▆█▇▇▇▃ #13 of 405 * · 27% of its score · 1 test

Beats GPT-6 Astra, GPT-6.1 Sol, Claude Sonnet 5.5 here, though they rank higher overall. Trails Muse Spark 1.2, Muse Spark 1.1, Claude Opus 4.6 here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
LMArena (text)1490.4 ✎▁▁▁▂▃▂▃▆▆▇█▃ beats 376/388avg 1262▇▇▇▇▇▇▇▇▇▇▇ 27%blunt

$ where to run it

Through an API

api · $3.00/M · $1.5 in / $7.5 out per 1M tokens · 1M context · ai.google.dev ↗ · price sourceUSD per 1M tokens, Gemini API Standard paid tier for gemini-4-hydrogen (output includes thinking tokens), flat across the context window. Cached input $0.075 (implicit caching only) and batch prices excluded. A free tier exists. Google official pricing page, observed 2026-11-14.

$ how the score was worked out · LMArena leaderboard dataset 77%, Terminal-Bench 4.0 official l… 21%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
LMArena leaderboard dataset2CC BY 4.0
Terminal-Bench 4.0 official leaderboard submissions1Apache-2.0
prior · pull toward the middle
weighted average= 79.1

$ improve this page

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json