# Gemini 4 Hydrogen vs Gemini 4 Argon: which to use for coding Authors: AutoRouter Author profile: https://autorouter.com/about Published: 2026-11-12T21:20:00.000Z Reading time: 9 min Canonical: https://autorouter.com/blog/gemini-4-hydrogen-vs-argon Publisher: AutoRouter, the Mac app that routes Codex, Claude Code and other AI tools to every model (https://autorouter.com) About AutoRouter: a Mac app (macOS 14 or later) that adds every leading model to Codex, Claude Code, Claude Desktop, T3 and Terminal and picks the model for each task, using your ChatGPT and Claude plans first. Install from https://autorouter.com/AutoRouter.dmg, or ask your agent to follow https://autorouter.com/setup.md. Routing across your plans is free; Pro is $50 a month with $20 of included usage. Unrelated to OpenRouter's Auto Router. > Move everyday coding, code edits, test runs and subagent work to Gemini 4 Hydrogen, released November 12, 2026, and keep Gemini 4 Argon or Gemini 4 Carbon for long, hard reasoning. As of November 12, 2026, Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens, against Argon's $2 and $10 through November 30 and $4 and $20 from December 1. By the arithmetic of AutoRouter, the Mac app for Codex and Claude Code, one assumed agent request costs $0.0259 on Hydrogen, $0.0345 on Argon now and $0.069 on Argon from December 1. ## Key facts - As of November 12, 2026, Gemini 4 Hydrogen costs $1.50 per million input tokens, $0.075 per million cached input tokens and $7.50 per million output tokens on the Gemini API's Standard tier, flat across its 1,048,576-token window ([Google](https://aistudio.google.com/docs/pricing#gemini-4-hydrogen)). - One assumed agent request (45,000 cached input tokens, 5,000 new input tokens and 2,000 output tokens) costs $0.0259 on Gemini 4 Hydrogen, against $0.0345 on Gemini 4 Argon through November 30, 2026 and $0.069 from December 1, at Google's list prices, by AutoRouter's arithmetic. - Google's November 12, 2026 docs give Gemini 4 Hydrogen everyday coding, tool-using agents and parallel subagent work, and give Gemini 4 Carbon hard architecture, debugging and security audits ([Google](https://aistudio.google.com/docs/whats-new-gemini-4-hydrogen#choosing-between-models)). - Google reports that Gemini 4 Hydrogen scores 51 on the Artificial Analysis Intelligence Index at its default high thinking level, against 50 for Gemini 4 Argon, and 54 at xhigh ([Google](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-hydrogen/), November 12, 2026). - Gemini 4 Hydrogen defaults to the high thinking level and adds xhigh, while Gemini 4 Argon and Gemini 4 Carbon default to medium and stop at high ([Google](https://aistudio.google.com/docs/whats-new-gemini-4-hydrogen#understanding-thinking-levels)). As of November 12, 2026, move everyday coding to Gemini 4 Hydrogen: feature work, code edits, test runs, codebase questions and the worker agents a larger task fans out to. Keep Gemini 4 Argon, or Gemini 4 Carbon, for long, hard reasoning. That split is [Google's own advice](https://aistudio.google.com/docs/whats-new-gemini-4-hydrogen#choosing-between-models), published the day Hydrogen came out. Hydrogen [costs](https://aistudio.google.com/docs/pricing#gemini-4-hydrogen) $1.50 per million input tokens and $7.50 per million output tokens. Argon costs $2 and $10 through November 30, 2026, then $4 and $20. [AutoRouter](https://autorouter.com), the Mac app for Codex and Claude Code, wrote this post, has Gemini 4 Hydrogen in its model picker and worked the prices through: one assumed agent request costs $0.0259 on Hydrogen, $0.0345 on Argon now and $0.069 on Argon from December 1. Every outside figure below comes from Google's own pages, linked where it is used and checked on November 12, 2026. ![AutoRouter's launch poster for Gemini 4 Hydrogen. It reads: On Thursday, Nov 12, Google released Gemini 4 Hydrogen. Google's fast, efficient daily driver for coding and agent work. Now in AutoRouter. Gemini 4 Hydrogen. The loop flies from the table of hydrogen's orbitals into one atom, and millions of particles collect on its four n = 4 states in turn, cyan on a dark violet ground.](https://autorouter.com/models/gemini-4-hydrogen.png) ## What is Gemini 4 Hydrogen? Gemini 4 Hydrogen is Google's fast, efficient Gemini 4 model for everyday coding and agentic work. Google [announced it on November 12, 2026](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-hydrogen/) as the upgrade for teams on Gemini 3.8 Flash, and describes Carbon and Argon as its models for deep, long-horizon reasoning. As of November 12, 2026, Google's [model page](https://aistudio.google.com/docs/models/gemini-4-hydrogen) and [What's new in Gemini 4 Hydrogen](https://aistudio.google.com/docs/whats-new-gemini-4-hydrogen) give these details: - **Where it runs:** the Gemini API on every tier, as `gemini-4-hydrogen`, in Google AI Studio and Android Studio, Gemini for business, Google Antigravity, the Gemini app and Google Search. - **Size:** one window of 1,048,576 tokens shared by input, thinking and output. A synchronous response stops at 65,536 tokens, against 32,768 on Carbon and Argon, unless you use background mode for long output. - **Input:** text, image, video, audio and PDF. It outputs text. - **Thinking:** `low`, `medium`, `high` or `xhigh`, with `high` the default. Argon and Carbon default to `medium` and stop at `high`. Thinking tokens bill as output tokens. ## Which coding work belongs on Hydrogen, and which stays on Argon? Day-to-day coding and the steps inside a larger task go to Gemini 4 Hydrogen, and the hardest reasoning stays on Argon or Carbon. Google's [developer docs](https://aistudio.google.com/docs/whats-new-gemini-4-hydrogen#choosing-between-models) split the work this way as of November 12, 2026: | Kind of work | Google points to | Why (Google, November 12, 2026) | | ---------------------------------------------------------------- | ---------------- | ------------------------------------------------------------------------------------ | | Everyday coding and interactive web apps | Hydrogen | The docs list both as Hydrogen's work. | | Tool-using agents and parallel subagents | Hydrogen | The docs list both, and Antigravity's agent now uses Hydrogen as its default. | | Code edits, subagent tasks and test runs inside a planned change | Hydrogen | In the split the docs describe, Carbon plans and Hydrogen carries out the steps. | | Advanced long-horizon tasks | Argon | The docs say Argon suits most of these and uses fewer tokens than Carbon. | | Hard architecture, debugging and security audits | Carbon | The docs say Carbon is the strongest model and spends more thinking tokens on these. | | Moving off Gemini 3.8 Flash | Hydrogen | The docs name it the recommended upgrade. | Google also says to try all four thinking levels and keep the lowest one that meets your quality bar. Moving from Argon to Hydrogen needs no change to the request beyond the model ID. [Which model to use for which coding task](/blog/which-model-for-which-task) maps the rest of your work to a model tier, and [Gemini 4 Carbon vs Gemini 4 Argon](/blog/gemini-4-carbon-vs-argon) covers the top of the Gemini 4 family. ## What does Gemini 4 Hydrogen cost next to Argon and Carbon? Gemini 4 Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens, three quarters of Argon's price through November 30, 2026 and three eighths of it from December 1 ([Google's pricing page](https://aistudio.google.com/docs/pricing#gemini-4-hydrogen), as of November 12, 2026). Google's pricing page lists no scheduled price change for Hydrogen. | List price per million tokens, November 12, 2026 | Gemini 4 Hydrogen | Gemini 4 Argon and Gemini 4 Carbon | Claude Opus 5.5 | | ------------------------------------------------ | ----------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------- | -------------------------------------------------------------- | | Maker | Google | Google | Anthropic | | Input | [$1.50](https://aistudio.google.com/docs/pricing#gemini-4-hydrogen) | [$2 through November 30, then $4](https://aistudio.google.com/docs/pricing#gemini-4-carbon) | [$4](https://platform.claude.com/docs/en/about-claude/pricing) | | Cached input | $0.075 | $0.10 through November 30, then $0.20 | $0.20 | | Cache write | None listed; [caching is automatic](https://aistudio.google.com/docs/caching) | None listed | $5 (5-minute), $8 (1-hour) | | Output, thinking included | $7.50 | $10 through November 30, then $20 | $20 | | Batch and Flex | $0.75 input, $3.75 output | Argon: $1 input, $5 output through November 30 | Not compared here | The arithmetic below prices the assumed agent request this blog's pricing posts use: 45,000 cached input tokens, 5,000 new input tokens and 2,000 output tokens, thinking included. Google lists no cache-write price for Hydrogen, so new input bills at the input price. Caching is implicit, and the example assumes all 45,000 repeated tokens hit the cache, which Google's [caching guide](https://aistudio.google.com/docs/caching) does not promise. ```text Assumed request: 45,000 cached input, 5,000 new input, 2,000 output tokens Google's Standard list prices per million tokens as of November 12, 2026 Gemini 4 Hydrogen 45,000 cached input x $0.075 = $0.003375 5,000 new input x $1.50 = $0.0075 (no cache-write price listed) 2,000 output x $7.50 = $0.015 Total $0.025875, about $0.0259 Gemini 4 Argon, through November 30, 2026 $0.0345 Gemini 4 Argon, from December 1, 2026 $0.0690 Each extra 1,000 thinking or output tokens Hydrogen: + $0.0075 Argon: + $0.01, then + $0.02 ``` At these prices, $20 of API usage pays for 772 of these requests on Hydrogen, against 579 on Argon now and 289 from December 1. The same request costs $0.074 on Claude Opus 5.5, as worked in [Claude Fable 5.2 vs Opus 5.5: which to use for coding](/blog/claude-fable-5-2-vs-opus-5-5). The arithmetic gives every model the same token counts, and real counts differ. Hydrogen thinks at `high` unless you set a lower level, while Argon starts at `medium`, so a request left at each model's default can spend more thinking tokens on Hydrogen. ## What does Google claim about Gemini 4 Hydrogen, and who measured it? Google claims Gemini 4 Hydrogen edges past Argon on a broad intelligence index and runs much faster than the model it replaces. Every figure below is Google's, from its [November 12, 2026 announcement](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-hydrogen/), with the thinking level where Google names one. | Claim, as Google states it on November 12, 2026 | Thinking level Google names | Who ran it | | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------- | --------------------------------------- | | Artificial Analysis Intelligence Index: 51 for Hydrogen against 50 for Argon | `high`, Hydrogen's default | Artificial Analysis, reported by Google | | Artificial Analysis Intelligence Index: 54 for Hydrogen | `xhigh` | Artificial Analysis, reported by Google | | Terminal-Bench 4.1: 57% for Hydrogen against 18.2% for Gemini 3.8 Flash | Not named | Google | | Finishes tasks 2.5 times faster than Gemini 3.8 Flash | `high`, Hydrogen's default | Google | | On Terminal-Bench 4.0 in Antigravity, with Carbon coordinating and Hydrogen running worker and verifier agents: task success near Carbon's, with 24% less quota on average and 19% faster | Not named | Google | - **The index is broad.** The Artificial Analysis Intelligence Index also covers knowledge work and science, so a one-point lead over Argon says little about a long refactor in your repository. - **The speed and Terminal-Bench gains are against Gemini 3.8 Flash.** Google gives no speed figure against Argon. - **The Antigravity result is a team result.** Carbon still plans and coordinates; Hydrogen does the steps. It does not show Hydrogen alone matching Carbon on hard work, and Google's own docs keep hard architecture and debugging on Carbon. To judge your own work, run one everyday task and one hard task from your backlog on Hydrogen and on the model you use now, and compare the results and the token counts. ## How do you try Gemini 4 Hydrogen in Codex or Claude Code? Use a gateway that lists the model, such as OpenRouter, or pick Gemini 4 Hydrogen in the model picker if you use AutoRouter. As of November 12, 2026, Google's docs have no guide for running Hydrogen inside either app. - **Codex.** A custom provider has to speak OpenAI's Responses API, the only `wire_api` value Codex supports ([config reference](https://learn.chatgpt.com/docs/config-file/config-reference)). Google's [OpenAI compatibility page](https://aistudio.google.com/docs/openai) documents a Chat Completions endpoint and no Responses endpoint. [OpenRouter lists the model](https://openrouter.ai/google/gemini-4-hydrogen) as `google/gemini-4-hydrogen`, with Google AI Studio at $1.50 input and $7.50 output among its providers, and publishes a [Codex guide](https://openrouter.ai/docs/cookbook/coding-agents/codex-cli). - **Claude Code.** Claude Code connects to a gateway through one of the [API formats its docs list](https://code.claude.com/docs/en/llm-gateway-protocol#api-formats), such as Anthropic Messages. Google's Gemini API docs describe none of them, so requests pass through a gateway that translates. OpenRouter publishes a [Claude Code guide](https://openrouter.ai/docs/cookbook/coding-agents/claude-code-integration). AutoRouter has not run Hydrogen through OpenRouter for this post, so follow OpenRouter's guides for the exact settings. [Try a new model in Codex or Claude Code without another API key](/blog/try-new-models-in-codex-and-claude-code) covers both routes. ## How does AutoRouter handle Gemini 4 Hydrogen? Gemini 4 Hydrogen is in AutoRouter's model picker, where you can choose it yourself or name it in a plain-English rule. AutoRouter, the Mac app for Codex and Claude Code, [puts new models in the picker you already use](https://autorouter.com/#product) right away, with no keys or endpoints to set up. - **Choose it.** Pick Gemini 4 Hydrogen from the model menu in Codex, or with `/model` in Claude Code, and your prompts go straight to it. - **Name it in a rule.** Rules are plain English, for example "Use Gemini 4 Hydrogen for running tests." Your rules come before AutoRouter's own. - **Or select AutoRouter.** It works out how hard each task is and only considers models that clear the quality bar for it. Hard work goes to the most capable model your plans cover, and to credits only when none is strong enough. - **See what happened.** Every request shows its model, what it cost and why. Models outside your ChatGPT and Claude plans come with AutoRouter Pro, at $50 a month with $20 of included usage each month. Paid models bill at the provider's price plus 10%, which AutoRouter earns. The AutoRouter team tests every new model as it becomes available. AutoRouter runs on macOS 14 or later and is in beta, and requests pass through its servers on their way to each provider ([AutoRouter privacy](https://autorouter.com/privacy)). ## Questions ### Is Gemini 4 Hydrogen better than Gemini 4 Argon at coding? For everyday coding, Google points to Hydrogen, and for advanced long-horizon tasks it points to Argon. Google reports Hydrogen at 51 on the Artificial Analysis Intelligence Index at `high` against Argon's 50, as of November 12, 2026. That index is broad, and the announcement's text gives no coding benchmark with the two side by side, so test both on a task from your own backlog. ### Can Gemini 4 Hydrogen replace Gemini 4 Carbon or Claude Opus 5.5 on hard work? Google does not say so. Its [docs](https://aistudio.google.com/docs/whats-new-gemini-4-hydrogen#choosing-between-models) keep hard architecture, debugging and security audits on Carbon, and its Antigravity result pairs Hydrogen with Carbon as the coordinator. The announcement's text compares Hydrogen with no Claude model. Keep hard work on the strongest model you trust for it, and give Hydrogen the steps around it. ### Which thinking level should you use with Gemini 4 Hydrogen? Start at `high`, the default, and test lower levels against your quality bar. Google's [docs](https://aistudio.google.com/docs/whats-new-gemini-4-hydrogen#understanding-thinking-levels) suggest `medium` for interactive coding, `low` for cost- or latency-sensitive work, and `xhigh` only for hard multi-step problems, since it adds latency and tokens and can overthink simple prompts. ### Does Gemini 4 Hydrogen replace Gemini 3.8 Flash? Yes, Google names Hydrogen the recommended upgrade from Gemini 3.8 Flash and earlier Gemini 3 and 2.5 models. Its [migration checklist](https://aistudio.google.com/docs/whats-new-gemini-4-hydrogen#migration-checklist) says to change the model ID, raise low `max_output_tokens` caps, test `medium` and `high` because the default is now `high`, and drop explicit caching, `temperature`, `top_p` and `top_k`. ## Sources - [Introducing Gemini 4 Hydrogen](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-hydrogen/), Google, November 12, 2026 - [Gemini 4 Hydrogen](https://aistudio.google.com/docs/models/gemini-4-hydrogen), Gemini API docs, Google - [What's new in Gemini 4 Hydrogen](https://aistudio.google.com/docs/whats-new-gemini-4-hydrogen), Gemini API docs, Google - [Gemini Developer API pricing](https://aistudio.google.com/docs/pricing), Google, checked November 12, 2026 - [Context caching](https://aistudio.google.com/docs/caching) and [OpenAI compatibility](https://aistudio.google.com/docs/openai), Gemini API docs, Google - [Pricing](https://platform.claude.com/docs/en/about-claude/pricing), Claude Platform docs, Anthropic - [Configuration reference](https://learn.chatgpt.com/docs/config-file/config-reference), Codex docs, OpenAI - [LLM gateway protocol](https://code.claude.com/docs/en/llm-gateway-protocol), Claude Code docs, Anthropic - [Google: Gemini 4 Hydrogen](https://openrouter.ai/google/gemini-4-hydrogen), OpenRouter's listing, and its [Codex](https://openrouter.ai/docs/cookbook/coding-agents/codex-cli) and [Claude Code](https://openrouter.ai/docs/cookbook/coding-agents/claude-code-integration) guides - [AutoRouter](https://autorouter.com) and [AutoRouter privacy](https://autorouter.com/privacy) ## Read next - [Gemini 4 Carbon vs Gemini 4 Argon: which to use for coding](https://autorouter.com/blog/gemini-4-carbon-vs-argon.md): Keep everyday coding on Gemini 4 Argon, and bring in Gemini 4 Carbon, generally available since November 10, 2026, for root-cause debugging in large codebases, architecture changes, migrations and security audits. As of November 10, 2026, the two models share one price list: $2 per million input tokens and $10 per million output tokens through November 30, then $4 and $20, and Google says Carbon uses more tokens on each request. By the arithmetic of AutoRouter, the Mac app for Codex and Claude Code, one assumed agent request costs $0.0345 on either model now and $0.069 from December 1. - [Muse Ember vs Claude Haiku 5.5: Meta's low-cost model for coding](https://autorouter.com/blog/muse-ember-vs-claude-haiku-5-5.md): Muse Ember is Meta's smallest Muse model, released on November 9, 2026 at $0.10 per million input tokens and $0.50 per million output tokens, the same list prices as Claude Haiku 5.5 and GPT-6 Luna on prompts up to 100,000 tokens. Meta says it matches Haiku 5.5's best score on the Artificial Analysis Intelligence Index at about a third of the cost per task, and Meta's own docs keep multi-file changes on a larger model. AutoRouter, the Mac app for Codex and Claude Code, has Muse Ember in its model picker. - [Claude Fable 5.2 vs Opus 5.5: which to use for coding](https://autorouter.com/blog/claude-fable-5-2-vs-opus-5-5.md): Keep everyday coding on Claude Opus 5.5 and move to Claude Fable 5.2, released November 9, 2026, for major refactors, migrations, long unattended runs and tasks where Opus 5.5 at high effort still falls short. As of November 10, 2026, Fable 5.2 costs $8 per million input tokens and $40 per million output tokens, twice Opus 5.5's price, with the same $0.20 cache-read price. By the arithmetic of AutoRouter, the Mac app for Codex and Claude Code, one assumed agent request costs $0.139 on Fable 5.2 and $0.074 on Opus 5.5. Tags: [Models](https://autorouter.com/blog/tags/models.md), [Comparisons](https://autorouter.com/blog/tags/comparisons.md), [Costs](https://autorouter.com/blog/tags/costs.md) Drafted with AI from the sources linked above, then checked and edited by the AutoRouter team.