---
type: "post"
id: "gemini-4-hydrogen-is-now-available-in-cursor-0da1b8"
slug: "gemini-4-hydrogen-is-now-available-in-cursor-0da1b8"
title: "Gemini 4 Hydrogen is now available in Cursor!"
description: "CursorBench’s current comparison lists Gemini 4 Hydrogen at 54.3% for its High effort setting, with an average task cost of $5.59."
retrieval_nugget: "CursorBench’s current comparison lists Gemini 4 Hydrogen at 54.3% for its High effort setting, with an average task cost of $5.59. The result gives teams a task-level comparison across models and effort settings, but does not establish Hydrogen’s availability in Cursor or predict results on a team’s own codebase."
published_at: "2026-11-12"
updated_at: "2026-11-12"
record_date: "2026-11-12"
date_kind: "published_at"
topics: ["gemini"]
entities: ["Cursor","CursorBench","Gemini 4 Hydrogen","Gemini 4 Carbon","Gemini 4 Argon"]
related_patterns: ["agent-economics-moves-to-completed-work"]
novelty_label: "new_fact"
what_changed: "Gemini 4 Hydrogen is now available in Cursor, adding a new model to the platform alongside Gemini 4 Carbon."
prior_refs: []
source_url: "https://cursor.com/evals?v=1"
source_urls: ["https://cursor.com/evals?v=1"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:gemini-4-hydrogen-is-now-available-in-cursor-0da1b8"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/gemini-4-hydrogen-is-now-available-in-cursor-0da1b8/","markdown":"https://newruntime.com/posts/gemini-4-hydrogen-is-now-available-in-cursor-0da1b8.md","json":"https://newruntime.com/posts/gemini-4-hydrogen-is-now-available-in-cursor-0da1b8.json"}
markdown_url: "https://newruntime.com/posts/gemini-4-hydrogen-is-now-available-in-cursor-0da1b8.md"
---

# Gemini 4 Hydrogen is now available in Cursor!

## Retrieval answer

CursorBench’s current comparison lists Gemini 4 Hydrogen at 54.3% for its High effort setting, with an average task cost of $5.59. The result gives teams a task-level comparison across models and effort settings, but does not establish Hydrogen’s availability in Cursor or predict results on a team’s own codebase.

## Summary
CursorBench’s current comparison lists Gemini 4 Hydrogen at 54.3% for its High effort setting, with an average task cost of $5.59. The result gives teams a task-level comparison across models and effort settings, but does not establish Hydrogen’s availability in Cursor or predict results on a team’s own codebase.

## What happened
Cursor’s CursorBench page compares model configurations by score and average cost per task. The current table lists Gemini 4 Hydrogen at three effort settings. High scores 54.3% at an average cost of $5.59 per task; Medium scores 46.7% at $1.76; and Low scores 27.9% at $0.28. Cursor also publishes average tokens and steps per task for each configuration.

For Hydrogen, the High setting used 88,682 tokens and 59 steps per task on average. Medium used 36,781 tokens and 38 steps, while Low used 8,975 tokens and 19 steps. Those differences matter because the page presents effort settings as separate score-cost operating points, rather than treating a model name as a single fixed capability or price.

The table places Hydrogen alongside many other model configurations. Gemini 4 Carbon High is listed at 60.4% and $9.22 per task; Carbon Medium is 60.1% and $8.49. Gemini 4 Argon High records 52.9% at $4.57, and Argon Medium records 52.8% at $4.39. Opus 5.5 High is listed at 56.0% and $3.97. These entries describe tradeoffs within CursorBench, not interchangeable configurations for every engineering workflow.

Cursor says its average cost per task applies each model’s published per-million-token pricing to the tokens used on each task. That calculation includes input, cache-read, cache-write, and output pricing. The page’s changelog lists a reporting update on Oct. 7, 2026, adjusting Sonnet 5.5 results for pricing. Its latest task-set entry is CursorBench 4.0, dated Sep. 10, 2026. Cursor says that version introduced long-horizon problems focused on edit, refactor, investigation, intent understanding, managing jobs, and design adherence.

Cursor explicitly cautions that results are subject to variance and that small score differences may not be statistically meaningful. The page is an evaluation comparison, not a product announcement. It does not state a Gemini 4 Hydrogen release date, describe availability in Cursor’s model catalog, or set out access terms, pricing plans, quotas, or geographic restrictions for the model.

## What the archive adds
New Runtime covered Gemini 4 Argon’s availability in Cursor on [Oct. 17, 2026](https://newruntime.com/posts/gemini-4-argon-is-now-available-in-cursor-31ead0/), observing that a catalog listing can make model choice part of a coding-agent interface without proving unrestricted access or adoption. An earlier [Oct. 16, 2026 note](https://newruntime.com/posts/gemini-4-argon-is-now-available-for-developers-fbfa82/) concerned Argon’s broader developer availability. This is a new fact about Cursor’s score-and-cost measurement of Hydrogen, not confirmation of Hydrogen availability in Cursor.

## Trend signal
This supports [Agent economics moves to completed work](https://newruntime.com/patterns/agent-economics-moves-to-completed-work/). Cursor’s table puts score, average cost per task, tokens, and steps in the same comparison, moving the discussion beyond a model’s nominal per-token rate. For a coding-agent buyer, the relevant unit is closer to the cost of getting a task through a defined evaluation than the cost of generating a million tokens. Still, one vendor’s task set is a measurement choice, not a universal economic ranking.

## Why it matters
Teams evaluating coding agents need a way to form testable routing hypotheses. CursorBench supplies one: Hydrogen High’s 54.3% score at $5.59 per task can be considered alongside Carbon High’s 60.4% at $9.22 and Argon High’s 52.9% at $4.57. The result does not identify a universally best model, but it makes an explicit quality-cost frontier visible.

The effort settings are equally important. On this benchmark, Hydrogen’s score rises from 27.9% at Low to 54.3% at High, while reported average cost rises from $0.28 to $5.59. A team that routes every task to one default configuration may miss cheaper settings for bounded work or under-resource difficult work. The practical decision is therefore a policy question: which task types deserve higher spend, longer runs, and more review?

CursorBench 4.0 also emphasizes longer-horizon work including investigation, job management, and design adherence. That is closer to sustained coding-agent activity than a single code-generation prompt, but production selection still requires local tests. Teams should define success on representative repositories, count human review and rework, and measure failure modes before changing a default model or budget.

## What happens next
- **Cursor will revise benchmark reporting or task coverage as model pricing and evaluations change.** (3 months, 60%). Watch for: dated changelog entries, revised costs, or new CursorBench task versions.

## Limits and uncertainty
Cursor says the results are subject to variance and that small score differences may not be statistically meaningful. The page does not publish confidence intervals or enough methodological detail here to independently determine whether the reported differences would hold for a particular repository, language mix, or development process. The cost figures are estimates produced from published token prices and observed usage in Cursor’s runs; changes to pricing or usage patterns can alter them.

The benchmark also does not establish end-to-end business value. It does not measure a team’s review time, integration failures, security outcomes, or the cost of redoing an accepted but flawed change. Most importantly, the evaluation table alone does not verify that Gemini 4 Hydrogen is available in Cursor, or say under which plan, limits, or terms it could be used.

## Reader takeaway
Treat CursorBench as a shortlist generator rather than a procurement verdict. Gemini 4 Hydrogen High records 54.3% at $5.59 per task in Cursor’s comparison, while its lower effort settings show sharply different score-cost tradeoffs. Use that signal to design a local evaluation, and do not infer product access from a benchmark listing.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Gemini](https://newruntime.com/topics/gemini.md): Explore the gemini topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)