Short answer
Google released Gemini 4 Hydrogen on November 12, 2026. The model ID is gemini-4-hydrogen. It is the fast, cheap, still-capable tier of the Gemini 4 family, and Google explicitly calls it the recommended upgrade from Gemini 3.8 Flash.
The numbers, all from Google:
- Price: $1.50 per million input tokens, $7.50 per million output tokens, with a free tier.
- Context: 1 million tokens, shared across input, thinking and output.
- Thinking levels:
low,medium,high(default) andxhigh—xhighis new. - Modality: text, image, video, audio and PDF in; text out.
gemini-4-hydrogen is not on TeamoRouter yet. Google shipped it, but it has not reached our catalog. Once it does, the key you already have works: change the model field to gemini-4-hydrogen and keep the base URL. To start today, use Gemini 4 Argon (gemini-4-argon) or Gemini 4 Carbon (gemini-4-carbon) — both are live on TeamoRouter.
What this model is for
Google's positioning is narrow and clear: bring Gemini 4 intelligence down to a price and speed that suits everyday coding and agentic work.
Hydrogen shares its architecture with Carbon and Argon — the same 1M unified context window, the same long-decode behaviour, the same built-in tools. The difference is how many thinking tokens it burns per turn, meaning how long it deliberates.
Google's own line is blunt: "Recommended upgrade from Gemini 3.8 Flash and earlier Gemini 3 and 2.5 models." Hydrogen takes over the Flash slot.
What Google announced
Everything below comes from Google's What's new in Gemini 4 Hydrogen and the model page.
- Cleaner code: Google says it writes better code and cuts fabricated claims by more than 5x compared with Gemini 3.8 Flash — by flagging missing context instead of guessing.
- Faster tool use: it batches read and search tool calls in parallel and adapts better when a tool call fails.
- Four thinking levels:
low,medium,high(default) andxhigh. Hydrogen is the model that introducesxhigh. - Built for multi-model work: it runs standalone, or alongside Carbon and Argon — Carbon plans and unblocks, Hydrogen executes in parallel.
- Stronger multimodal reasoning: more accurate on tables, charts, dense PDFs, images, audio and video, plus agentic video understanding that inspects specific frames.
- 1M context and long decode: input, thinking and output share a 1,048,576-token window. Synchronous output defaults to 65,536 tokens and can run to 1M with long decode.
- Now the default for Managed Agents: the Antigravity agent in Gemini Managed Agents uses Hydrogen by default.
Some things it cannot do, per the official model page: no image generation, no Live API, no audio generation. Context caching is implicit only — explicit caching is not supported.
Pricing
Official Google prices, per 1 million tokens, USD:
| Free tier | Standard (paid) | |
|---|---|---|
| Input | Free | $1.50 |
| Output (including thinking tokens) | Free | $7.50 |
| Context caching | Free | $0.075 (implicit only) |
The other consumption options are priced differently:
| Option | Input | Output | Caching |
|---|---|---|---|
| Standard | $1.50 | $7.50 | $0.075 |
| Batch | $0.75 | $3.75 | $0.0375 |
| Flex | $0.75 | $3.75 | $0.0375 |
| Priority | $2.70 | $13.50 | $0.135 |
Two things worth flagging:
- Thinking and tool instructions bill as tokens. Google states on the pricing page that every token the model processes is counted, including thinking tokens and the internal tool system instructions and function declarations used with grounding.
- Explicit caching is not available, so repeated input only pays off when implicit caching hits. When it misses, you pay full input price.
Google Search grounding is not available on the free tier. On the paid tier you get 5,000 free search requests a month (shared across all Gemini 3 and newer models), then $14 per 1,000 requests with a daily cap of 1,500. Separately, free-tier traffic is used to improve Google's products; paid-tier traffic is not.
Choosing a thinking level
This is the most consequential change in Hydrogen, and Google's guidance is specific:
| Level | Token use | When to use |
|---|---|---|
low |
Fewest tokens, lowest latency | Cost- and latency-sensitive work: fast chat, high-throughput processing. Still capable on agentic tasks |
medium |
Lower token use, faster | Interactive coding, live UI prototyping, responsive chat, latency-sensitive tool calls |
high (default) |
Balanced | Most tasks: complex coding, agentic execution, tool calling, multimodal knowledge work |
xhigh |
Most tokens, deepest reasoning | Only for hard multi-step coding and agentic problems where high falls short, and where you can afford the extra latency and tokens |
Google's advice: start with high for most complex tasks, and reserve xhigh for problems where high is not enough. xhigh over-thinks simple prompts, which makes them slower and more expensive. Run your evaluations at all four levels with a high max_output_tokens so the cap does not truncate answers, then pick the lowest level that meets your quality bar.
Hydrogen vs Carbon vs Argon
All three share the 1M context window and built-in tools. Pick by reasoning depth, latency and cost:
| Gemini 4 Hydrogen | Gemini 4 Carbon | Gemini 4 Argon | |
|---|---|---|---|
| Official price (per 1M in / out) | $1.50 / $7.50 | Intro $2 / $10 | Intro $2 / $10 |
| Position | Fast and cheap, still handles complex agentic flows | Strongest, deepest reasoning | Handles most long-horizon tasks with fewer tokens than Carbon |
| Best for | Everyday coding, interactive web apps, tool-using agents, multimodal documents, parallel subagents | Hard architecture, debugging, science, security audits | Long-horizon work where you want to spend fewer tokens |
| Available on TeamoRouter | Not yet | Yes | Yes |
Google also suggests a pairing worth stealing: have Carbon plan system interfaces and clear hard roadblocks, then hand code edits, parallel subagent tasks and test runs to Hydrogen. The expensive model only makes decisions; the cheap one does the work.
Calling it on TeamoRouter (once it lands)
Not available yet. When gemini-4-hydrogen enters our catalog, your key and base URL stay put — only model changes:
from openai import OpenAI
client = OpenAI(base_url="https://api.teamorouter.com/v1", api_key="YOUR_KEY")
resp = client.chat.completions.create(
model="gemini-4-hydrogen",
messages=[{"role": "user", "content": "Explain what a context window is in one sentence."}],
)
print(resp.choices[0].message.content)
The native Gemini format uses the same host:
curl https://api.teamorouter.com/v1beta/models/gemini-4-hydrogen:generateContent \
-H "x-goog-api-key: YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"contents": [{"parts": [{"text": "Explain what a context window is in one sentence."}]}]}'
gemini-4-hydrogen is not on TeamoRouter yet. Once it is live, the same key and model: gemini-4-hydrogen is all you change; until then, use gemini-4-argon.
What to use today
- Gemini 4 Argon (
gemini-4-argon): same Gemini 4 generation, live on TeamoRouter. Intro price $2 / $10 through November 30, 2026, then $4 / $20. One tier pricier than Hydrogen, but it handles heavier work. See Gemini 4 Argon API. - Gemini 4 Carbon (
gemini-4-carbon): same intro price, built for the hardest reasoning and cybersecurity. See Gemini 4 Carbon API. - Gemini 3.8 Flash (
gemini-3.8-flash): cheaper for now ($0.75 / $3.75 through December 31, 2026), good for high-volume, simple calls.
How to pay
No subscription: top up your balance in the console from $5 and each request is billed by the tokens it uses. Cards are supported, and so are USDT, USDC and ETH. The same key also calls Claude and GPT models.
New accounts start with a $0 balance — top up before your first request. Top up from $5. If a request returns 402 (400 on the OpenAI format) with "wallet balance is insufficient", your balance is empty — your base URL and key are fine. Top up and retry.
FAQ
Is Gemini 4 Hydrogen out?
Yes. Google made it generally available in the Gemini API on November 12, 2026 under the model ID gemini-4-hydrogen, on both the free and paid tiers.
How much does Gemini 4 Hydrogen cost?
$1.50 per million input tokens and $7.50 per million output tokens, with context caching at $0.075 (implicit only). There is a free tier. Batch and Flex are half price; Priority is $2.70 / $13.50.
Is it the same model as Gemini 4 Flash?
Yes. Leakers on X had called this Flash-tier Gemini 4 model "Hydrogen", and Google used that name officially. The model ID is gemini-4-hydrogen, not gemini-4-flash. Our earlier preheat page used gemini-4-flash and has been moved to gemini-4-hydrogen.
Can it generate images?
No. The official model page lists image generation, audio generation and Live API as unsupported.
Does it support the Batch API?
Yes, at half price: $0.75 input, $3.75 output. Flex and Priority inference are supported too. Explicit caching is not — only implicit.
Can I call gemini-4-hydrogen on TeamoRouter now?
Not yet. Google shipped it, but it has not entered our catalog. Once it does, the same key and a new model ID is all it takes. Use Gemini 4 Argon or Gemini 4 Carbon for now.
Once it is live, is it just a model-name swap?
Yes. It runs on the standard generateContent path like Gemini 4 Argon and Carbon, not on a separate endpoint like Gemini Decision. Same key, same base URL, only model changes.
Related reading
- Using Gemini 4 Hydrogen from China and Russia: what you need without a Google account or card
- Gemini 4 Hydrogen vs Carbon vs Argon: the full comparison
- Gemini 4 Argon API: the same generation, available on TeamoRouter today
- Gemini API changelog: the official source