Strengths
- It scores 54 on the Artificial Analysis index v5.0.1 at the highest reasoning setting (xhigh) and 51 at the default. The older Gemini 4 Argon scores 50 and costs more in the API.
- Google says it finishes tasks 2.5 times faster than Gemini 3.8 Flash at the default setting, scores 57% against 18.2% on Terminal-Bench 4.1 and makes more than five times fewer unsupported claims.
- It writes quickly: about 174 tokens per second according to Artificial Analysis, against 110 for Gemini 4 Neon.
- You can try it on the free API tier. The $1.50 / $7.50 price covers the whole one-million-token window, and batch processing costs half.
- It takes in about a million tokens of text, images, video, audio and PDFs at once, and in long decode mode, which runs in the background, it can write very long answers. It can operate a computer and check answers against Google Search and Maps.
Weaknesses
- At the highest reasoning setting it is very wordy and slow to start. It produced 304 million output tokens on the Artificial Analysis test against a 77 million median and sent the first part of its answer after about 47 seconds.
- In Artificial Analysis's independent Terminal-Bench 4.0 run it scores 53.5%, below Claude Sonnet 5.5 (63.6%) and GPT-6.2 Sol (60.1%).
- Until the end of 2026 it costs twice as much as Gemini 3.8 Flash at its introductory $0.75 / $3.75. From 1 January 2027 the prices are the same.
- Google's model card says it refuses without good reason more often than 3.8 Flash, mainly on prompts with images. Google has not published the full safety report yet.
- In the free Gemini app you cannot pick it for good. Google says free users get it through automatic model selection when it suits the prompt, but most answers still come from Flash-Lite.
Best for
- everyday coding and debugging
- AI agents that work through many steps quickly, including as helpers to the stronger Carbon
- work with documents, spreadsheets and charts
- trying out the Gemini API for free
Not ideal for
- the hardest long tasks, where Gemini 4 Neon or Gemini 4 Carbon is the better pick
- high volumes of simple tasks, where Gemini 3.5 Flash-Lite is cheaper
- apps that need an instant answer at high reasoning settings
How to use it in Europe
Since 12 November 2026 you can use it in Europe through the Gemini API and Google AI Studio, on both the free and paid tiers. In the Gemini app it comes with Google AI Plus, Pro and Ultra, while free users only get it through automatic model selection. Google Antigravity and Google Sheets include it from AI Pro, and companies find it in Gemini Enterprise. On Gemini Enterprise Agent Platform (formerly Vertex AI) it also runs in the eu region, so data can stay in the EU. Under its terms, Google does not train on API data from users in the EEA, even on the free tier.
Price compared with similar models
Benchmark results
| Benchmark | Score | Source |
|---|---|---|
| Artificial Analysis Intelligence Index v5.0.1 (xhigh effort) | 54 | artificialanalysis.ai ↗, November 2026 |
| Terminal-Bench 4.0 (Artificial Analysis, xhigh) | 53.5% | artificialanalysis.ai ↗, November 2026 |
| Artificial Analysis output speed (xhigh) | 174 tokens/s | artificialanalysis.ai ↗, November 2026 |
| Terminal-Bench 4.1, cheating-adjusted | 57.0% (Gemini 3.8 Flash: 18.2%) | deepmind.google ↗ (vendor-reported), November 2026 |
| HLE-Diamond | 58.8% | deepmind.google ↗ (vendor-reported), November 2026 |
Benchmarks cover only a narrow slice of ability and vendors often report them at their best settings. Treat them as a rough guide.
More models from this vendor
- Gemini 4 Carbon: Google's most capable model, available since 10 November 2026 in the paid API, for businesses and in Google AI Ultra. It scores 60 on the independent index.
- Gemini 4 Neon: Google's powerful model between Carbon and Hydrogen, out since 18 November 2026. In the app it comes with Google AI Pro in place of Argon.
- Gemini 3.5 Flash-Lite: A very fast, low-cost Google model for large volumes of simple tasks, such as pulling data out of documents.
- Gemma 4 31B: Google's largest open model, which you can run on your own computer or server when data must not leave the company.
- Gemma 4 26B A4B: Google's faster open model for your own server. Close to the larger Gemma 4 31B in quality, but quicker to respond.
Similar models from other vendors
- Claude Sonnet 5.5 (Anthropic): Claude's balanced model for everyday work and coding. It is faster than Opus 5.5 and costs half as much.
- Command A+ (Cohere): An open business model from Canada's Cohere that officially supports Slovak and Czech and runs on your own server.
- Muse Spark 1.3 (Meta): Meta's balanced model with a 1-million-token window and a fair price. The Muse Spark line also powers the Meta AI assistant.
- Muse Spark Open (Meta): Meta's large open model that a company can run on its own servers, so data never leaves its infrastructure.
- Llama 4 Maverick (Meta): The larger model in the newest Llama generation, from April 2025. Cheap through providers, but its license excludes people and companies in the EU.
- Nemotron 3.5 Super VL (NVIDIA): NVIDIA's open model for business AI agents that understands text, images and video and is very fast and cheap to run.
Sources
- Google: Introducing Gemini 4 Hydrogen ↗
- Google: Gemini 4 Hydrogen for developers ↗
- Google DeepMind: Gemini 4 Hydrogen model card ↗
- Gemini API: Gemini 4 Hydrogen ↗
- Gemini API: What's new in Gemini 4 Hydrogen ↗
- Gemini API pricing ↗
- Gemini API changelog ↗
- Gemini Enterprise Agent Platform: Gemini 4 Hydrogen ↗
- Gemini subscriptions (SK) ↗
- Google: Search upgrade with Gemini 4 Hydrogen ↗
- Artificial Analysis: Gemini 4 Hydrogen ↗
- Artificial Analysis: Terminal-Bench 4.0 ↗
- Gemini API Additional Terms of Service ↗
- Artificial Analysis: GPT-6.2 Sol vs Claude Opus 5.5 ↗
- Artificial Analysis: Gemini 4 Neon ↗