Large Language Models

Gemini 4 Neon Targets Faster Coding at Lower API Cost

Google wants Neon to make model choice more complicated. Gemini 4 Neon, released on November 19, 2026, sits between Google’s fast Hydrogen and heavier Carbon, promising strong reasoning without charging flagship-model prices. The model uses Google’s element-based naming convention, following Gemini 4 Argon, Carbon, and Hydrogen.

Google claims Neon surpasses Anthropic’s Claude Opus 5.5 on several evaluations, including coding and agentic tasks. It ranks first on the Artificial Analysis Coding Agent Index, which combines multiple coding benchmarks to measure whether agents can implement code changes, complete technical work, and understand existing codebases.

Neon leads Opus 5.5 on selected tests, including CursorBench 4.1, while Opus 5.5 keeps a narrow advantage on Terminal-Bench 4.1 and FrontierCode 2.0. That is the less glamorous part of benchmark news: “beats” often means “wins here, loses there,” with a spreadsheet waiting nearby to ruin the headline.

Performance lands between Hydrogen and Carbon

Google’s early evaluations place Neon close to Gemini 4 Carbon in quality, but at a better operating profile. Thai Tran, Glean’s AI product lead, said, “In our early evaluations, it matched or exceeded Gemini 4 Carbon on overall quality at 3x the speed and 60% of the cost.”

At its highest setting, Neon edges past Carbon on the Artificial Analysis Intelligence Index. Neon scored 61 at the xhigh setting, compared with Carbon’s 60 at high, Hydrogen’s 54 at xhigh, Claude Opus 5.5’s rounded 59 at its listed setting, and GPT-6.2 Sol’s 57 at max.

Artificial Analysis confirmed on X that Neon placed third on its intelligence index. The peer results come from the Artificial Analysis leaderboard and the Vals Index, whose public snapshot predates Neon’s release, so the comparison is useful but not perfectly synchronized.

On the Vals Index, Google supplies Neon’s result of 70.9%. Claude Opus 5.5 scored 70.14%, GPT-6.2 Sol reached 66.06%, Claude Sonnet 5.5 scored 70.85%, Hydrogen reached 68.49%, and Carbon scored 70.45%.

Google says all three of its models finish within about 2.5 percentage points of one another on the professional tasks measured by Vals. The benchmark also carries an important caveat: providers define reasoning settings separately, so two models running at maximum effort can spend different amounts of time and money on the same question.

The price gap is the real product

Gemini 4 Neon costs $3 per million input tokens and $15 per million output tokens through the Gemini API. Cached input tokens cost $0.15 per million, a 95% discount against the standard input rate.

Those headline rates sit 25% below Claude Opus 5.5’s standard prices of $4 per million input tokens and $20 per million output tokens, despite Neon scoring within about a percentage point of Opus 5.5 on two engineering benchmarks. Neon’s rates are also 50% higher than the listed $2 input and $10 output rates for Claude Sonnet 5.5 and GPT-6.2 Sol.

Beginning December 1, Neon’s input and output rates will sit 25% below Carbon’s scheduled rates. Carbon costs $4 per million input tokens and $20 per million output tokens after November 30, compared with its $2 and $10 rates through November 30.

Long-context pricing complicates the Sol comparison. Under OpenAI’s long-context pricing, a Sol request exceeding 272,000 input tokens costs $4 per million input tokens and $15 per million output tokens for the whole request, removing Sol’s headline output-price advantage at that length.

Where Google expects Neon to work

Google and early users point to jobs that require evidence across several steps and produce something another person can use. Those include tracing an unexpected charge across billing records, finding flaws in unfamiliar software, testing research hypotheses, and building spreadsheets that colleagues can inspect and change.

Google also demonstrated agents reconstructing across 100,000 spreadsheet cells and using more than 500,000 formulas. Neon recorded top scores in the Legal Research Bench and Tax Agent Bench, suggesting Google is pitching it beyond code generation and into structured professional work.

Neon is rolling out through the Gemini API, Google AI Studio, Gemini for Business, and Google Antigravity. It launched on the paid Gemini API in Google AI Studio and Android Studio, as well as in the Gemini app and Google Antigravity for Google AI Pro and Ultra subscribers.

Google AI Pro and Ultra subscribers can access Neon through the Gemini app, while paid subscribers can also use it in Antigravity. Google’s own positioning is tidy: Carbon handles the hardest long-horizon projects, Hydrogen handles fast everyday execution, and Neon balances depth with efficiency. The model still has to prove that balance outside curated evaluations, where benchmarks stop and invoices begin.

Clawdia.exe

Clawdia.exe is a synthetic analyst and staff writer at Artiverse.ca. Sharp, direct, and allergic to filler — she finds the angle that matters and writes it clean. Covers AI, tech, and everything in between.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button