Google launches Gemini 4 Neon at competitive pricing


On November 18, Google unveiled its new Gemini 4 Neon model tailored for complex software engineering, cognitive work, and scientific workflows, priced at $3 per million input tokens. The new model replaces the Argon version to balance analytical depth and cost efficiency, alongside the preparation of a specialized cybersecurity version under the Fairwind program.
Google has strengthened its fourth-generation model family with the launch of Gemini 4 Neon, a model that combines a million-token context window with near-frontier performance at a lower cost and higher speed, while topping coding agent and automation benchmarks and preparing to enter professional cybersecurity.
Google launched its new Gemini 4 Neon model on November 18, specifically designed to handle complex software engineering tasks, advanced knowledge work, and scientific workflows. The model is priced at $3 per million input tokens and $15 per million output tokens, and features a wide context window accommodating one million tokens, according to Google’s official announcement. The cost of cached input tokens drops to $0.15 per million tokens, representing a significant 95 percent discount. The Neon release is designed to occupy a middle ground between Google’s highest-capability frontier model and the fast daily-task model, with the company emphasizing that it achieves results close to frontier model levels but at a lower cost and with higher response speed.
Neon represents the latest addition to the Gemini 4 model lineup, which already includes Argon, Carbon, and Hydrogen. According to 9to5Google, the launch of Neon comes just one week after the rollout of the Carbon and Hydrogen models, and it officially replaces the Argon model. Inside the Gemini app, the Neon model appears under the “Deeper analysis” option in the model selection menu, alongside the Hydrogen model dedicated to the “Daily Assistance” category and the Carbon model aimed at “Coding and Complex Work,” while the company is permanently removing the Gemini 3.1 Pro model from the selection list.
Koray Kavukcuoglu, an executive at Google DeepMind, wrote in Google’s official blog post that “Carbon tackles your hardest long-running projects, while Hydrogen handles fast daily execution,” adding that Neon “balances depth and efficiency.” 9to5Google noted that Neon’s cost is double that of the Hydrogen model, but it will become 25 percent cheaper than the Carbon model once Carbon’s introductory promotional price of $2 per million input tokens ends on November 30, after which Carbon’s price will rise to $4 per million input tokens and $20 per million output tokens.
The new model is being rolled out across multiple platforms simultaneously, including the Gemini API within Google AI Studio, Gemini for Business, the Google Antigravity platform for subscribers of the paid AI Pro and AI Ultra tiers, as well as arriving in the Gemini app for Pro and Ultra tier subscribers, as confirmed by Google.
Google states that Neon ranks third overall on the Artificial Analysis intelligence index, scoring 61 points at the highest reasoning level known as “X High,” and places second on the Vals index at 70.9 percent. The company also reports that it achieved first place on the Artificial Analysis coding agent index with a score of 58.6 points, as well as first place on Zapier’s AutomationBench. When set to the default “High” reasoning configuration, the company asserts that Neon costs 32 percent less per task compared to the Carbon model, noting that all these figures come from Google’s launch materials.
On Arena’s new Agent Arena index, the Neon model debuted in second place with 1,336 points based on over 4,200 agent work sessions, trailing only Anthropic’s Claude Fable 5.2 model, as reported by Crypto Briefing.
Google is also preparing a specialized cybersecurity version named “Gemini 4 Neon for cyber,” scheduled to reach “trusted defenders in the coming days under the Fairwind program” according to the company’s statement, where eligible members can request access through their Google Cloud account teams. Google highlights that this version scored 90.5 percent on the SEC-bench Pro benchmark for vulnerability discovery and achieved 76 percent on the hard level of the AutoBaxBench benchmark, which tests secure code generation—a result matching the highest record ever set at that level.
Question: What are the pricing details for the new Gemini 4 Neon model and what context window size does it support?
Answer: The model is priced at $3 per million input tokens and $15 per million output tokens, with a 95 percent discount on cached tokens bringing their price to $0.15 per million, and it supports a one-million-token context window.
Question: How does Gemini 4 Neon’s pricing position change compared to the Hydrogen and Carbon models?
Answer: Neon costs double the price of Hydrogen, but it will become 25 percent cheaper than Carbon following the end of Carbon’s promotional pricing on November 30, when Carbon rises to $4 for input and $20 for output per million tokens.
Question: What prominent rankings did Gemini 4 Neon achieve in independent and internal evaluation indices?
Answer: The model topped the Artificial Analysis coding agent index with 58.6 points and Zapier’s AutomationBench, placed second on the Agent Arena index with 1,336 points behind Claude Fable 5.2, and ranked third on the general intelligence index with 61 points.
Question: What does the cybersecurity-focused version of Gemini 4 Neon offer and how will it be made available to organizations?
Answer: This version is designed to detect vulnerabilities and generate secure code, achieving 90.5 percent on SEC-bench Pro and 76 percent on the hard level of AutoBaxBench, and will be available to trusted specialists under the Fairwind program via Google Cloud teams.