Gemini 4 Hydrogen is Google's AI model, released 12/11/2026. Gemini 4 Hydrogen is Google DeepMind's fast, lower-cost Gemini 4 model and the successor to Gemini 3.8 Flash. Google's table compares it with Claude Sonnet 5.5 and GPT-6.1 Terra: it leads on most rows but trails Sonnet 5.5 on Terminal-Bench 4.1, OSWorld 2.1 and Terminal-Bench Science. Results use the highest thinking setting unless stated.
Visit Gemini 4 HydrogenAI-Powered
Leverages advanced AI technology to deliver cutting-edge capabilities and results.
Fast & Efficient
Optimized performance ensures quick results without compromising on quality.
Purpose-Built
Specifically designed for llms tasks and workflows.
How does Gemini 4 Hydrogen score on independent benchmarks?
Gemini 4 Hydrogen ranks #8 of 28 models on the Artificial Analysis Intelligence Index (v5.0.1) with a score of 53.9, as measured on 22 November 2026.
Where does Gemini 4 Hydrogen rank against other models?
How good is Gemini 4 Hydrogen at agentic coding?
How does Gemini 4 Hydrogen do on individual tests?
| Evaluation (run by Artificial Analysis) | Gemini 4 Hydrogen | Best of any model |
|---|---|---|
| Terminal-Bench 4.0 | 53.5% | 67.2% |
| FrontierCode 2.0 | 39% | 48.8% |
| Humanity's Last Exam | n/a | 74.3% |
| GDPval-AA (Elo) | n/a | 1866 |
How has Gemini 4 Hydrogen's score changed over time?
We have tracked Gemini 4 Hydrogen on Intelligence Index v5.0.1 since 22/11/2026. A point is added each day the score, coding index or price changes, so this chart fills in over time.
What does the Gemini 4 Hydrogen system card report?
Gemini 4 Hydrogen is Google DeepMind's fast, lower-cost Gemini 4 model and the successor to Gemini 3.8 Flash. Google's table compares it with Claude Sonnet 5.5 and GPT-6.1 Terra: it leads on most rows but trails Sonnet 5.5 on Terminal-Bench 4.1, OSWorld 2.1 and Terminal-Bench Science. Results use the highest thinking setting unless stated.
These are Google DeepMind's own published results (self-reported, with the vendor's harness and settings), so they are not directly comparable with other labs' numbers. Sources: Gemini 4 Hydrogen Model Card (undefined/11/2026), Gemini 4 Hydrogen Model Evaluation (approach, methodology and results) (undefined/11/2026), Introducing Gemini 4 Hydrogen (12/11/2026). Last checked 22/11/2026.
Coding
| Benchmark | Gemini 4 Hydrogen | Claude Sonnet 5.5 | GPT-6.1 Terra | Gemini 4 Argon |
|---|---|---|---|---|
| SWE-sweepsubtask-weighted reward; Antigravity-CLI 1.4.8 harness, safety filters on; self-computed | 13.5% | 3.1% | 5.8% | n/a |
| Terminal-Bench 4.1cheating-adjusted accuracy; high effort, mini-swe-agent vision harness, average over 5 attempts; self-computed | 57% | 63.9% | n/a | n/a |
Agents & tool use
| Benchmark | Gemini 4 Hydrogen | Claude Sonnet 5.5 | GPT-6.1 Terra | Gemini 4 Argon |
|---|---|---|---|---|
| Agents' Last Exam v1.1score; 122-task set, Antigravity-CLI 1.4.8, 5-hour window, safety filters on; self-computed | 71.6% | 60.4% | 69.8% | n/a |
| AutomationBench-AAArtificial Analysis leaderboard | 76.1% | 71.8% | 60.2% | n/a |
| Vending-Bench 2final bank balance; Andon Labs leaderboard | 14549 USD | 11165 | 9107 | n/a |
| OSWorld 2.1offline subset, max of 3 runs, partial score; self-computed | 77.6% | 83.9% | 74.2% | n/a |
| Vals Tax Agent Benchrank #1 overall per launch post | 52.3% | n/a | n/a | n/a |
| Vals Legal Research Benchtied #1 per launch post | 55.8% | n/a | n/a | n/a |
Multimodal
| Benchmark | Gemini 4 Hydrogen | Claude Sonnet 5.5 | GPT-6.1 Terra | Gemini 4 Argon |
|---|---|---|---|---|
| MKA (Multimodal Knowledge Agent)Climb WSR accuracy; self-computed | 53.4% | 52% | 43.8% | n/a |
| Chartographyno tools; self-computed following the Surge leaderboard | 66.6% | 61.6% | 63.2% | n/a |
Knowledge & science
| Benchmark | Gemini 4 Hydrogen | Claude Sonnet 5.5 | GPT-6.1 Terra | Gemini 4 Argon |
|---|---|---|---|---|
| Terminal-Bench Science 0.1mini-swe-agent vision harness, average over 5 attempts; self-computed | 51.4% | 59.9% | 58.3% | n/a |
Reasoning & maths
| Benchmark | Gemini 4 Hydrogen | Claude Sonnet 5.5 | GPT-6.1 Terra | Gemini 4 Argon |
|---|---|---|---|---|
| HLE-DiamondArtificial Analysis leaderboard | 58.8% | 53.3% | 47.3% | n/a |
Other
| Benchmark | Gemini 4 Hydrogen | Claude Sonnet 5.5 | GPT-6.1 Terra | Gemini 4 Argon |
|---|---|---|---|---|
| Artificial Analysis Intelligence Indexdefault high thinking; per launch post | 51 points | n/a | n/a | 50 |
| Artificial Analysis Intelligence Index (xhigh)xhigh thinking; per launch post | 54 points | n/a | n/a | n/a |
What did safety testing find for Gemini 4 Hydrogen?
Safety level: Frontier Safety Framework (April 2026): CBRN and Cybersecurity CCLs not reached after targeted testing, but the CCL alert thresholds were reached in both domains
- No Material Capability Change versus Gemini 4 Argon, so a full critical capability suite was not required; targeted CBRN and cyber assessments were run because Argon and Carbon could not be ruled out for the biological CBRN CCL and Cybersecurity CCL.
- Hydrogen reaches the CCL alert thresholds for both CBRN and cyber while not reaching the CCLs themselves; mitigations continue.
- Loss of Control: control, monitoring and model security mitigations are applied as if the Stealth and Situational Awareness TCL had been reached, as a conservative assumption.
- Versus Gemini 3.8 Flash: text safety +0.80pp, multilingual +1.10pp, image-to-text +7.00pp, tone -0.65pp, unjustified refusals +5.57pp (higher, which is worse).
- Launch post (company-reported): overreach in agentic tasks reduced by up to 81%, misaligned actions and task shortcutting cut 12x, and 2.3x more resilient on Gray Swan IPI.
Independent figures: Artificial Analysis, refreshed daily on our AI model benchmarks page. Full analysis: our Gemini 4 Hydrogen deep dive.
Google Model Timeline
1M tokens (up to 64K default synchronous output, up to 1M via background long decode) context
1M tokens (up to 64K default synchronous output, up to 1M via background long decode) context
1M tokens (combined input and output) context
1M tokens context
1,049k tokens context
66k tokens context
1,049k tokens context
33k tokens context
1,049k tokens context
1,049k tokens context
1,049k tokens context
8k tokens context
1,049k tokens context
1,049k tokens context
1,049k tokens context
8k tokens context
33k tokens context
1,049k tokens context
33k tokens context
96k tokens context
33k tokens context
131k tokens context
131k tokens context
96k tokens context
1,049k tokens context
1,049k tokens context
1,049k tokens context
8k tokens context
8k tokens context
Specifications
Related Tools
Google Veo
Google Veo is DeepMind's flagship text-to-video model, producing high-resolution, cinematic clips with strong prompt adherence and, in newer versions, synchronised audio. It is integrated across Google's creative and cloud products.
NotebookLM (Gemini Notebook)
Renamed Gemini Notebook in July 2026, this is Google's AI research and note-taking assistant that grounds its answers in your own uploaded sources. Known for source-grounded summaries, Audio Overviews, and, on paid tiers, a new sandboxed cloud computer that writes code and exports charts, slides and spreadsheets.
Claude Opus 5.5
Claude Opus 5.5, launched 22 September 2026, is Anthropic's current Opus-line flagship, replacing Opus 5. It performs at roughly the level of Claude Fable 5.1 on most work while costing 40% less to run than Opus 5, at $4 input / $20 output per million tokens (both 20% down, cache reads cut 60% to $0.20). It leads on Terminal-Bench 4.0, GDPval-AA v2.1 and Humanity's Last Exam, and is the first Opus model to ship with Fable-5.1-class cybersecurity, biology and distillation safeguards.
