Gemini 4 Hydrogen vs Claude Sonnet 5.5: Price and Free Tier

Gemini 4 Hydrogen costs $1.50 input and $7.50 output per million tokens, 25% less than Claude Sonnet 5.5 at $2/$10, and it is the only Gemini 4 model on the free API tier. Boards checked 20 November 2026 put Sonnet 5.5 ahead on web design (Arena WebDev 1,771 vs 1,701) and Hydrogen ahead on cost.

Viết với sự hỗ trợ của AI: bản nháp hướng dẫn này được soạn cùng Claude Sonnet 5.5, và mỗi dữ kiện đều dẫn link đến nguồn của nó.

Gemini 4 Hydrogen is the cheaper model and Claude Sonnet 5.5 the higher-ranked one for web pages. Hydrogen costs $1.50 input and $7.50 output per million tokens against Sonnet's $2 and $10 (Gemini API pricing), and it is free in Google AI Studio. On 20 November 2026 Sonnet 5.5 ranked #3 on Arena WebDev and Hydrogen #14.

For all four Gemini 4 models see which Gemini 4 model to use; for Google's strongest one see Gemini 4 Carbon vs Claude Fable 5.2.

How much does Gemini 4 Hydrogen cost compared with Claude Sonnet 5.5?

Gemini 4 Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens, thinking included, at one flat price across its 1,048,576-token window (Gemini API pricing, checked 2026-11-20). Claude Sonnet 5.5 costs $2 and $10 (Anthropic), so Hydrogen is 25% cheaper on both. Batch and Flex halve Hydrogen's price.

Gemini 4 Hydrogen spec and price, checked 2026-11-20
ItemGemini 4 HydrogenSource
Model IDgemini-4-hydrogen, stable only, no preview or -latest aliasModel page
Context window1,048,576 tokens shared by input, thinking and outputWhat's new
Max output65,536 tokens per normal request; about 1M only with long decodeWhat's new
Thinking levelslow, medium, high (default), xhighThinking
Knowledge cutoffMarch 2026Model card
Price per 1M tokens$1.50 input, $7.50 output, $0.075 cachedPricing
Batch, Flex, Priority$0.75 / $3.75 for Batch and Flex; $2.70 / $13.50 for PriorityPricing
Free tierYes, standard input and output; not Batch or FlexPricing
CachingImplicit only, from 4,096 input tokens, 95% offCaching
Vertex AISame on the global endpoint; $1.65 / $8.25 regionalGoogle Cloud pricing

Unlike Argon and Carbon, Hydrogen has no introductory price, so nothing changes for it on 1 December 2026. Google's choosing-a-model page names Hydrogen the default Gemini 4 model.

Is Gemini 4 Hydrogen free?

Yes. As of 20 November 2026 Hydrogen is the only one of the four Gemini 4 models on the free Gemini API tier, where its standard input and output are free of charge, and it is free in Google AI Studio (Gemini API pricing). Google publishes no request or token limits for that tier.

  • Gemini API: the free tier covers standard input and output, not Batch or Flex, and free-tier prompts are used to improve Google's products.
  • AI Studio: Google lists Hydrogen on every plan, including Free (AI Studio plans), and forum staff wrote on 18 November 2026 that it is free "in Playground or Build mode" (Google AI forum).
  • Rate limits: no per-minute or per-day numbers for any Gemini 4 model; they appear only in the signed-in AI Studio dashboard (rate limits).

Which Gemini app plans include Gemini 4 Hydrogen?

Every plan, including Free, as of 20 November 2026 (Gemini Apps Help). Free users reach Hydrogen through Auto, which picks between Hydrogen and an unnamed lightweight model, and in Tasks. Choosing Hydrogen by hand needs AI Plus at $4.99 a month or higher, and a hand-picked model runs at Low thinking unless you change it.

Gemini app models by plan (US prices), checked 2026-11-20
PlanPriceChats / TasksUsageContext
Free$0Auto / HydrogenStandard32k
AI Plus$4.99 a monthAuto, Hydrogen / Hydrogen2x128k
AI Pro$19.99 a monthAuto, Hydrogen, Neon / Hydrogen, Neon4x1M
AI Ultra$99.99 or $199.99Auto, Hydrogen, Neon, Carbon / same three5x or 20x AI Pro1M

Usage limits are set per plan, not per model; the limit rules and the full plan-by-model table are in the Gemini 4 hub guide.

Tasks, which runs on Hydrogen, began rolling out free on 17 November 2026 in six countries (release notes). Hydrogen also runs in Google Search for all users (Search post).

What does Google say Gemini 4 Hydrogen is good at?

Fast everyday coding and agent work. At launch on 12 November 2026 Google called Hydrogen "our daily driver for agentic workflows" and said it "generates more functional, feature-complete layouts in fewer prompts" (launch post). Its model card says Hydrogen is adapted from Gemini 4 Carbon.

The skill, gemini-frontend-design in google-gemini/demolab, also installs in Claude Code; its origin, motion rules and Google's preference figures are covered in the Carbon guide.

Google's model card compares Hydrogen only with Claude Sonnet 5.5 and GPT-6.1 Terra (evaluation PDF). Sonnet 5.5 leads three of the ten rows: Terminal-Bench 4.1 (63.9% vs 57.0%, Hydrogen at high), OSWorld-2.1 and Terminal-Bench-Science. Hydrogen leads the other seven, including SWE-sweep and Agents' Last Exam, both run in Google's own Antigravity harness.

How do Hydrogen and Sonnet 5.5 score on independent benchmarks?

Independent boards put Hydrogen below Claude Sonnet 5.5 on quality and well below it on cost. Artificial Analysis scores Hydrogen 54 at xhigh and 51 at its default high, against 55 for Sonnet 5.5 at max, and Arena WebDev ranks Hydrogen #14 at 1,701 against Sonnet's #3 at 1,771, checked 2026-11-20 (Artificial Analysis, Arena).

Hydrogen and Sonnet 5.5 on independent boards, checked 2026-11-20 (thinking level in brackets)
BoardGemini 4 HydrogenClaude Sonnet 5.5Source
AA Intelligence Index v5.0.154 (xhigh), #26 of 95; 51 (high)55 (max)AA Hydrogen
AA cost per index task$3.16 (xhigh); about $1.87 (high, motif's calculation from AA token counts)$10.02 (max)AA leaderboard
AA output speedAbout 164 tokens per second (xhigh)About 144 tokens per second (max)AA Hydrogen, AA Sonnet 5.5
Arena WebDev#14 of 162, 1,701 ±14 (high)#3, 1,771 (xHigh); #13, 1,716 (High)Arena
Arena WebDev cost per task (P50)$1.36 (high)$3.04 (xHigh); $1.29 (High)Arena
Agent Arena (18 November)#10 of 68 (spread 5 to 12), 1,271, preliminary; Google's blog said #5#5, 1,296 (max)Agent Arena
Vals Index v2.268.49%, #7 (xhigh)70.85%, #3Vals
Vals Vibe Code Bench v1.187.78%, #1592.39%, #3Vals
Official Terminal-Bench 4.156.4% (high, Antigravity CLI)63.9% (max, Claude Code)tbench.ai
Design Arena 3D Design#12, 1,396 (624 battles)#13, 1,392 (1,542 battles)Design Arena
AA-3D (Three.js scenes)Not listed yet#4, 1,204 (max)AA-3D

Sonnet 5.5 leads on Arena WebDev, Vibe Code Bench, the Vals Index, Agent Arena and Terminal-Bench 4.1. Hydrogen costs less per task on Artificial Analysis and against Sonnet at xHigh on Arena, and it edges Sonnet by 4 points on Design Arena's 3D board, too small a gap to call.

Thinking level matters. Sonnet 5.5 at High is #13 on Arena WebDev (1,716) at $1.29 a task, just above Hydrogen on score and just below it on cost. In Arena's Frontend category Hydrogen is #14 (1,726) and Sonnet 5.5 #3 (1,811).

Is Hydrogen or Sonnet 5.5 better for animation and web pages?

Sonnet 5.5 ranks higher; Hydrogen costs less and is free to try. On 20 November 2026 Sonnet 5.5 ranked #3 on both Arena WebDev and Vibe Code Bench at $2/$10 per million tokens, while Hydrogen ranked #14 and #15 at $1.50/$7.50. No same-prompt animation benchmark covers both models yet.

Gemini 4 Hydrogen vs Claude Sonnet 5.5 for front-end and motion work, checked 2026-11-20
Gemini 4 HydrogenClaude Sonnet 5.5
Model IDgemini-4-hydrogenclaude-sonnet-5-5
API price per 1M tokens$1.50 / $7.50$2 / $10
Context window1,048,576 tokens, flat price1M tokens, standard rates throughout
Max output per request65,536; about 1M with long decode128K; 300K via Batch API beta
Free useAPI free tier, AI Studio, Gemini appClaude Free plan includes Sonnet
In AntigravityPaid AI Pro and UltraPaid AI Pro, Ultra and Enterprise

What does one animation page cost with Gemini 4 Hydrogen?

About nine and a half cents. A request with 3,000 input tokens and 12,000 output tokens costs $0.0945 on Hydrogen at list price, against $0.126 on Claude Sonnet 5.5 and $0.189 on Neon, and half that with Batch or Flex (Gemini API pricing, checked 2026-11-20). Thinking tokens bill as output, so higher levels cost more.

motif ran a small test on 20 November 2026, Hydrogen at thinking level high, costed at paid list price (the free-tier key billed $0). Two runs of Google's quickstart prompt and one of a motion title-card prompt each returned a single-file HTML page for $0.089 to $0.097 in about 2.5 to 8 minutes. Four fixed motion prompts followed:

motif's Hydrogen motion test, thinking high, one sample each, run 2026-11-20 (cost at list price)
PromptTimeCostWhat we saw
SVG pelican on a bicycle, SMIL or CSS only8.5 min, 5 attempts$0.24Met every spec check; stopped under reduced motion
CSS landing hero with a scroll-driven effect2.5 min$0.08Met every spec check; stopped under reduced motion
GSAP 3.12.5 ScrollTrigger product story2.6 min$0.08Met every spec check; one label lagged in a scrolled frame
Three.js 0.160.0 lighthouse at dusk2.5 min$0.09Met every spec check; beam cone not visible at 2.6 s

Each output is an HTML file; to turn one into a video, see HTML animation to MP4. For the Claude side, see animated websites with Claude Code and Opus 5.5 vs Sonnet 5.5 for animation.

Where can you use Gemini 4 Hydrogen for coding?

Hydrogen runs free in Google AI Studio, in Antigravity for paid AI Pro and Ultra subscribers, in GitHub Copilot from the Pro plan up, in Cursor on Premium, Super and Ultra, and on OpenRouter as google/gemini-4-hydrogen. Gemini CLI has no Gemini 4 model registered, checked 2026-11-20.

  • Antigravity: default High plus an Extra (xhigh) level (models). Orchestrator can plan with Neon, Carbon or Argon while Hydrogen implements, and Antigravity CLI v1.5.8 added Hydrogen for Gemini API key users on 13 November 2026 (changelog).
  • Gemini CLI: the README says "Not optimized for Gemini 4+" (Gemini CLI); /model set rejects Gemini 4 IDs, and support is requested in issue #30385.
  • GitHub Copilot: Pro and above; Free and Student get only Auto, which has no Gemini 4 model (Copilot docs).
  • Cursor: $1.50 input, $0.075 cache read, $7.50 output; not on Free or Plus (Cursor docs).
  • OpenRouter: google/gemini-4-hydrogen at $1.50/$7.50, an OpenRouter slug, not a Google ID (OpenRouter).
  • Vertex AI: generally available since 12 November 2026 on global, us and eu endpoints (Vertex).

Has motif archived any Gemini 4 Hydrogen animations?

Not yet: none of the 28 Gemini 4 posts in motif's archive on 20 November 2026 names Hydrogen, Neon or Carbon, and its 22 same-prompt Argon runs against Claude are covered in the Argon guide.

Two of those runs pit Claude Sonnet 5.5 against Argon: Avenox's voxel Hagia Sophia time-lapse in one HTML file, both at high effort, and nevian's voxel Porsche in Three.js, which judges Sonnet's better than Argon's. More on Claude Sonnet 5.5 motions.

Each credit is the creator's claim, and 26 of the 28 entries credit Claude, so the sample is one-sided.

Should you use Gemini 4 Hydrogen or Neon?

Use Hydrogen for drafts, high volume and free work, and Neon when quality matters more than cost. Neon costs $3/$15 per million tokens, twice Hydrogen's price, has no free tier, and ranks #5 on Arena WebDev at 1,763 against Hydrogen's #14 at 1,701, checked 2026-11-20 (Gemini API pricing, Arena WebDev).

  • Speed and cost: at xhigh, Artificial Analysis measures about 164 output tokens per second for Hydrogen and 110 for Neon, and $3.16 against $6.44 per index task (AA Neon).
  • Web apps: Vibe Code Bench has Neon at 93.70% (#2) and Hydrogen at 87.78% (#15).
  • Access: Neon needs AI Pro in the Gemini app and a paid API tier.
  • Replacement: Google places Neon between Hydrogen and Carbon, and Hydrogen has no shutdown date (deprecations).

More in Gemini 4 Neon vs Claude Opus 5.5.

Questions

Is Gemini 4 Hydrogen free?

Yes, with limits Google does not publish. As of 20 November 2026 Hydrogen is the only Gemini 4 model on the free Gemini API tier, it is free in Google AI Studio's Playground and Build mode, and free Gemini app users reach it through Auto and in Tasks.

How much does Gemini 4 Hydrogen cost?

$1.50 per million input tokens and $7.50 per million output tokens, thinking included, 25% less than Claude Sonnet 5.5's $2/$10. Batch and Flex cost $0.75 and $3.75. Unlike Argon and Carbon, the price does not change on 1 December 2026 (checked 20 November 2026).

What is the Gemini 4 Hydrogen model ID and context window?

The ID is gemini-4-hydrogen, a stable version with no preview alias. It has a 1,048,576-token window shared by input, thinking and output, returns up to 65,536 tokens per normal request, and has a March 2026 knowledge cutoff, per Google's docs checked 20 November 2026.

Which Gemini app plan do I need for Gemini 4 Hydrogen?

Free is enough: as of 20 November 2026 every Gemini app plan includes Hydrogen, and free users reach it through Auto and in Tasks. Picking it by hand needs AI Plus at $4.99 a month in the US, or higher. Context is 32k tokens on Free, 128k on AI Plus and 1M on AI Pro and Ultra.

Which thinking level should I use for animation code?

Start with the API default, high. Google's docs suggest medium for live UI prototyping and offer xhigh for harder work, but no Google page recommends a level for 3D or animation. In the Gemini app a hand-picked model defaults to Low, so choose High there (checked 20 November 2026).

Is Gemini 4 Hydrogen better than Claude Sonnet 5.5 for web design?

Not on independent boards. On 20 November 2026 Arena WebDev ranked Sonnet 5.5 (xHigh) #3 at 1,771 and Hydrogen (high) #14 at 1,701, and Vibe Code Bench had Sonnet 5.5 at 92.39% and Hydrogen at 87.78%. Hydrogen costs 25% less per token and is free in AI Studio.

Is Gemini 4 Hydrogen faster than Claude Sonnet 5.5?

On token speed, yes. On 20 November 2026 Artificial Analysis measured about 164 output tokens per second for Hydrogen at xhigh and about 144 for Sonnet 5.5 at max. In one creator's same-task test Hydrogen finished in about 15 minutes against Sonnet's 32, a single unverified report.

Can Gemini 4 Hydrogen make animations or videos?

It writes animation code, not video. Hydrogen outputs text only, so an animation from it is HTML, SVG, CSS, GSAP or Three.js code that you render to MP4 afterwards. Google's "idea to finished video" demo of 13 November 2026 relied on a custom skill around the model.

Is Gemini 4 Hydrogen replacing Gemini 3.8 Flash?

Google recommends Hydrogen as the upgrade, but 3.8 Flash is not deprecated and its requests are not rerouted. On the Gemini API, 3.8 Flash costs $0.75/$3.75 per million tokens through 31 December 2026, then $1.50/$7.50, the same as Hydrogen. Hydrogen itself has no shutdown date (checked 20 November 2026).

Nguồn

  1. Google: Introducing Gemini 4 Hydrogen
  2. Gemini API docs: Pricing
  3. Gemini API docs: Gemini 4 Hydrogen model page
  4. Gemini API docs: What's new in Gemini 4 Hydrogen
  5. Gemini API docs: Thinking
  6. Gemini API docs: Rate limits
  7. Gemini API docs: Caching
  8. Gemini API docs: Choosing a model
  9. Gemini API docs: Deprecations
  10. Google AI Studio docs: Google AI plans
  11. Gemini API docs: What's new in Gemini 4 Neon (quickstart prompt)
  12. Google DeepMind: Gemini 4 Hydrogen model card
  13. Google DeepMind: Gemini 4 Hydrogen evaluation methodology (PDF)
  14. Google Cloud: Model Garden pricing
  15. Google Cloud: Gemini 4 Hydrogen on Vertex AI
  16. Gemini Apps Help: About Gemini models
  17. Gemini Apps Help: Usage limits and plans
  18. Google: Gemini subscriptions (US)
  19. Gemini app release notes
  20. Google: Gemini 4 Hydrogen in Search
  21. Google Antigravity: Gemini 4 Hydrogen in Antigravity
  22. Google Antigravity docs: Models
  23. Google Antigravity docs: Changelog
  24. GitHub: google-gemini/demolab
  25. GitHub: google-gemini/gemini-cli
  26. GitHub: gemini-cli issue #30385
  27. Google AI Developers Forum: free model access
  28. Claude Platform Docs: Claude Sonnet 5.5
  29. Claude Platform Docs: Pricing
  30. Claude Platform Docs: Models overview
  31. Claude: Plans and pricing
  32. Artificial Analysis: Gemini 4 Hydrogen
  33. Artificial Analysis: Models leaderboard
  34. Artificial Analysis: Gemini 4 Neon
  35. Artificial Analysis: Claude Sonnet 5.5
  36. Artificial Analysis: AA-3D
  37. Arena: WebDev leaderboard
  38. Arena: Gemini 4 Hydrogen on WebDev
  39. Arena: WebDev Frontend
  40. Arena: Gemini 4 Hydrogen on Agent Arena
  41. Vals AI: Vals Index
  42. Vals AI: Vibe Code Bench
  43. Terminal-Bench leaderboard
  44. Design Arena: 3D Design leaderboard
  45. GitHub Docs: Copilot supported models
  46. Cursor docs: Models and pricing
  47. OpenRouter: Models API
  48. @marcthecreatorr on X: ship test
Tất cả hướng dẫn