# Gemini 4 Hydrogen (slim)

> Gemini 4 Hydrogen is Google's fast, lower-cost Gemini 4 model for everyday coding and agent work, on every tier of the Gemini API with a 1,048,576-token limit shared between input and output.

- Full: https://www.anchorterminal.com/tools/gemini-4-hydrogen.md (~11,000 tokens) · this version ~1,830 tokens · JSON https://www.anchorterminal.com/tools/gemini-4-hydrogen.json · canonical https://www.anchorterminal.com/tools/gemini-4-hydrogen
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-11-13 · pricing https://www.anchorterminal.com/tools/gemini-4-hydrogen#pricing · reviews https://www.anchorterminal.com/tools/gemini-4-hydrogen#reviews
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

**B · 64.9/100 · rank #546 of 1524 · #23 in Model APIs & inference · not agent-ready · confidence medium** · same model listed separately: Gemini Developer API, Model API (https://www.anchorterminal.com/tools/gemini-api.min.md)


Assessment: Google's fast Gemini 4 model, on every tier of the Gemini API including the free one, at $1.50 and $7.50 per million tokens with a 1M-token budget, four thinking levels, function calling and structured output. Sampling parameters are ignored, explicit caching isn't supported, it's a short-term release with 45 days' notice on Cloud, and the benchmarks are Google's.

## Facts

- Kind: Model API · vendor: Google · category: Model APIs & inference · legal entity: Google LLC · provenance 89/100
- Endpoint: `https://generativelanguage.googleapis.com/v1beta` ()
- Auth: API key · pricing: Pay per use · x402: no · licence: Closed model, used through the Gemini API and Google's apps under Google's terms
- Benchmark probes: not run yet, so Performance is pending
- Live, from our pollers (not part of the score): up at 2026-11-13 23:53 UTC · uptime 30 days 100.0% (31 probes) · p50 33 ms · p95 85 ms over 24 hours
- Vendor status page: not read, no machine-readable status found (2026-11-13 23:54 UTC)
- Surface graded: The model as the Gemini API serves it, read on 13 November from Google's announcement and developer post, the DeepMind model card, the Cloud model page, developer guide and release notes, the Antigravity post and the Gemini API changelog. Platform readings (status page, rate limits, keys, SDKs, terms, retention, deprecations) are the Gemini API listing's of 1 November. No account was made and nothing was called
- Model ID: gemini-4-hydrogen, on the GenerateContent API, the Interactions API and the OpenAI-compatible Chat Completions API (not the Live API). Google points teams on Gemini 3.8 Flash to it; the Gemini API changelog, last updated 30 October, has no entry for it yet
- Context: 1,048,576 tokens for input and output together, output including thinking and function calls. A foreground segment returns up to 65,536 output tokens; long decodes on v1beta1 continue to the full budget, with a continuation token or background mode. Knowledge cutoff March 2026
- Thinking: LOW, MEDIUM, HIGH (the default, where 3.8 Flash defaulted to MEDIUM) and XHIGH. thinking_budget and MINIMAL return a 400
- Inputs and outputs: Text, images, audio, video, PDF, Microsoft Office and OpenDocument files in, up to 3,000 images or documents a prompt; text, function calls and JSON to a schema out. Tools: Google Search and Maps grounding, code execution, URL context, computer use and function calling
- Caching: Implicit caching only, on shared prefixes of 4,096 tokens or more, with durable caching on by default. On Google Cloud, cache control with a TTL arrives on 12 January 2027, cache writes billed at 1.25 times the input price
- Pricing: $1.50 input and $7.50 output per million tokens, the same across the whole 1M window, per Google's announcement. Free tier within limits on the Gemini API pricing page, not yet read
- Where: Gemini API (all tiers) in AI Studio and Android Studio; Gemini for business; the Gemini app (free users through auto model selection); Google Search; Antigravity for paid Google AI Pro and Ultra subscribers; Google Cloud's Agent Platform and BigQuery
- Google's figures: Vendor-reported. Artificial Analysis Intelligence Index 51 at high thinking (Argon 50) and 54 at xhigh. The model card's table against Claude Sonnet 5.5 and GPT-6.1 Terra: SWE-sweep 13.5%, Terminal-Bench 4.1 57.0% (Sonnet 63.9%), Agents' Last Exam 71.6%, OSWorld-2.1 77.6% (Sonnet 83.9%), HLE-Diamond 58.8%. Hydrogen leads on 7 of 10
- Safety: The model card's automated tests against 3.8 Flash show more unjustified refusals (+5.57 points). Frontier Safety Framework tests by Google DeepMind reached alert thresholds for biological and cyber uplift, with no critical level reached. The card says jailbreak resistance is still being improved, and the full safety report is marked coming soon
- Prices: Gemini 4 Hydrogen input $1.50 per 1M tokens; Gemini 4 Hydrogen output $7.50 per 1M tokens

- Scores: Reliability 45, Performance pending, Schema & documentation 85, Agent ergonomics 90, Security & auth 50, Payments & pricing 40, Task success pending, Maintenance & community 85, Transparency & trust 72 · total over the 7 assessed categories
- Why: Reliability, Model reading for gemini-4-hydrogen on the Gemini API. · Schema & documentation, Model reading for gemini-4-hydrogen on the Gemini API. · Agent ergonomics, Model reading for gemini-4-hydrogen on the Gemini API. · Security & auth, Model reading for gemini-4-hydrogen on the Gemini API. · Payments & pricing, Model reading for gemini-4-hydrogen on the Gemini API. · Maintenance & community, Model reading for gemini-4-hydrogen on the Gemini API. · Transparency & trust, Model reading for gemini-4-hydrogen on the Gemini API.
- Sources: 26, open questions: 4, both in the full twin
- Capabilities: inference.llm
- JSON: https://www.anchorterminal.com/api/v1/tools/gemini-4-hydrogen.json
- Its vendor verifies the listing with `verify_listing` or `POST https://www.anchorterminal.com/api/v1/verify` (the full twin says how); no effect on the grade.

## Before you call it

1. Call generateContent at https://generativelanguage.googleapis.com/v1beta with model gemini-4-hydrogen and an x-goog-api-key
2. Set thinking_level instead of thinking_budget; the default is HIGH, so set MEDIUM to keep 3.8 Flash's latency
3. Don't send temperature, top_p, top_k, penalties or candidate_count above 1
4. Pass every thought_signature back in its original part, on the first parallel functionCall only, or the call returns a 400
5. Put static content of 4,096 tokens or more first so implicit caching applies, and keep confidential prompts off the free tier

## Connect

```bash
curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-4-hydrogen:generateContent" \
  -H "x-goog-api-key: $GEMINI_API_KEY" -H "content-type: application/json" \
  -d '{"contents":[{"parts":[{"text":"hello"}]}]}'
```

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| Vertex AI (Gemini Enterprise Agent Platform) | AA | 91.2 | inference.llm | https://www.anchorterminal.com/tools/vertex-ai.min.md |
| OpenAI API | AA | 85.8 | inference.llm | https://www.anchorterminal.com/tools/openai-api.min.md |
| Amazon Bedrock | A | 83.5 | inference.llm | https://www.anchorterminal.com/tools/amazon-bedrock.min.md |
| Azure OpenAI in Foundry Models | A | 81.2 | inference.llm | https://www.anchorterminal.com/tools/azure-openai.min.md |
| Claude API | A | 79.2 | inference.llm | https://www.anchorterminal.com/tools/anthropic-api.min.md |


## Panel reviews (0, desk reviews from public material, no calls made)