Google launches Gemini 4 Neon for software workflows, claims 3x Carbon's speed
Google is on a launch spree in the artificial intelligence space. Just days after introducing Gemini 4 Carbon and rolling out Gemini 4 Hydrogen on November 12, the tech giant unveiled a brand-new model called Gemini 4 Neon on November 18, 2026.
Designed to sit right in the middle of the lineup, Neon brings high-end reasoning to everyday complex tasks without making users wait or pay a fortune. With this release, Google has replaced Gemini 4 Argon in the Gemini app and officially retired the older Gemini 3.1 Pro.advertisement❮❯ Read Full StoryWhere Gemini 4 Neon fits inIf Google’s periodic table naming is getting confusing, the breakdown is actually quite simple. Gemini 4 Hydrogen handles quick daily tasks, while Carbon is the heavy-duty flagship built for massive projects.
Neon bridges the gap, offering Carbon-level smarts for coding, research, and data analysis. According to Google, Neon runs roughly three times faster than Carbon while cutting overall costs by about 60 per cent.
It supports text, audio, image, and video inputs, and lets developers pick from four thinking levels, Low, Medium, High, and Extra High to balance speed and depth.Strong benchmark scores across coding and science
Independent testing backs up much of Google’s hype, even if real-world speed gains vary slightly. Evaluation firm Artificial Analysis gave Neon a score of 61 on its Intelligence Index at the Extra High setting.
That places Neon third overall, just behind Anthropic’s Claude Fable 5.2 at 65 and OpenAI’s GPT-6.2 Astra at 62, while edging past Carbon’s score of 60.
In science and maths, it posted 70.5 per cent on Terminal-Bench Science, 84 per cent on RiemannBench, and 71.3 per cent on HLE-Diamond.Built for business and cybersecurity
Neon also shines in office and security tasks. It scored 70.9 per cent on the Vals Index, topped both the Legal Research Bench and Tax Agent Bench, and hit 73.5 per cent on Chartography for reading charts.
On Zapier’s AutomationBench, it scored 56.3 per cent, and earned $16,843 in Andon Labs’ Vending-Bench 2 simulation. On the safety front, Neon scored 90.5 per cent on SEC-bench Pro for spotting bugs, 70.7 per cent on Collinear’s CWE-bench v2 for fixing them, 76 per cent on AutoBaxBench’s hard tier, and 51.26 per cent on WebExploitBench v1.2.
It also recorded a zero per cent attack success rate on Gray Swan’s prompt injection test, alongside a 0.09 per cent misaligned action rate and a 99.2 per cent honesty score on Google’s safety evaluations.Pricing catch and how to use it
Google has priced Neon at $3 per million input tokens, 15 cents for cached inputs, and $15 per million output tokens. This puts Neon between OpenAI’s GPT-6.2 Sol at $2 and Anthropic’s Claude Opus 5.5 at $4, meaning businesses pay 50 per cent more on inputs than they would for Sol.
There is also a temporary catch: Carbon is discounted to $2 input and $10 output until November 30, making Neon pricier this month before Carbon jumps to $4 and $20 on December 1.
Regular users with Google AI Pro or Ultra plans can try Neon today in the Gemini app by selecting Deeper analysis, while developers can access it via Google AI Studio, Android Studio, Gemini for business, and Google Antigravity, with a cybersecurity version coming soon to Fairwind Program members.- EndsAlso Read | Google is changing Gemini AI plans from October 9, here is what it means for youAlso Read | Adobe and Google team up to bring Gemini AI to Photoshop, PDF editing to GmailAlso Read | Satya Nadella sounds alarm over AI cyberattacks, says rushed security fixes are causing more crashes
I am a journalist covering the ever-evolving consumer tech landscape. From smartphones to laptops, I track the latest drops and decode the data to see if the newest gadgets are worth the hype.
