Google Launches Gemini 4 Neon, a Faster Mid-Tier Model for Coding and Knowledge Work

Gemini 4 Neon abstract illustration

Google Launches Gemini 4 Neon, a Faster Mid-Tier Model for Coding and Knowledge Work

Google DeepMind launched Gemini 4 Neon on November 18, 2026, a new model that sits between the fast Gemini 4 Hydrogen and the flagship Gemini 4 Carbon. Google describes Neon as bringing frontier-level intelligence with fast end-to-end task completion across complex software engineering, knowledge work, and science workflows. The announcement was written by Koray Kavukcuoglu, senior vice president at Google DeepMind and chief AI architect at Google.

A fourth model in a fast-growing family

Neon is the fourth model in the Gemini 4 family, arriving less than a week after Hydrogen, which launched on November 12. Google positions the lineup in three tiers. Hydrogen handles fast everyday execution, Carbon tackles the hardest long-horizon projects, and Neon balances depth and efficiency for complex, high-frequency work.

The launch also reshuffles the Gemini app. Neon replaces Argon in the model picker, and the older Gemini 3.1 Pro model is being retired. Paid subscribers now see four choices in the picker: Auto, which adapts to the task; Everyday help, powered by Gemini 4 Hydrogen; Deeper analysis, powered by Gemini 4 Neon; and Code and complex work, powered by Gemini 4 Carbon.

Key details: pricing, context, and thinking levels

Neon is priced at $3 per million input tokens and $15 per million output tokens through the Gemini API. Cached input costs $0.15 per million tokens, a 95 percent discount on the standard input rate. The model has a 1 million-token context window and accepts text, images, audio, and video as input, so it can synthesize across spreadsheets, PDFs, charts, audio, and video into structured deliverables.

Neon offers four configurable thinking levels, Low, Medium, High, and Extra High, letting teams trade intelligence against token spend. At its default high thinking level, Google says Neon delivers a high intelligence score at a 32 percent lower cost per task than Carbon.

Pricing context matters for anyone budgeting API use. Argon and Carbon currently cost $2 per million input tokens and $10 per million output tokens under promotional pricing that runs until November 30. After that, their rates rise to $4 and $20, which leaves Neon about 25 percent cheaper than Carbon from December 1.

Benchmarks: strong scores, with a caveat

Google reports that Neon ranks third overall with a score of 61 on the Artificial Analysis Intelligence Index at its Extra High thinking level, just behind larger and slower models. It also ranks first overall on the Artificial Analysis Coding Agent Index with a score of 58.6, second on the Vals Index at 70.9 percent, and first on both the Vals Legal Research Bench and the Tax Agent Bench.

The announcement lists further results across enterprise and scientific work, including first place on Zapier AutomationBench and Andon Labs Vending-Bench 2, and best-in-class scores on Terminal-Bench Science, RiemannBench, and HLE-Diamond. These are Google-reported figures from benchmarks Google chose to highlight, and independent testing will show whether the Carbon-level claim holds up on real work.

Availability

Neon began rolling out on November 18 in the Gemini API in Google AI Studio on the paid tier, and in Android Studio. Enterprises can use it through Gemini for business, via the API or the model dropdown, and in Google Antigravity. Consumers on Google AI Pro or Ultra plans can select it from the model dropdown in the Gemini app, and paid Pro and Ultra subscribers can also use it in Antigravity.

A separate cyber variant of Neon, tuned for defensive security work such as finding and patching code vulnerabilities, is rolling out to eligible members of Google Fairwind Program over the coming days. Google says this variant produces more secure code than Carbon across easy, medium, and hard challenges on the AutoBaxBench evaluation.

Why it matters

Neon fills the gap Google left between its cheapest and most powerful models. Developers and knowledge workers who found Hydrogen too light and Carbon too slow or too expensive now have a middle option that Google claims matches Carbon on coding, reasoning, and agentic tasks while running faster and costing less per task. For teams running long agent workflows, where a model reads many files, calls tools, and works through multi-step jobs, that combination of speed and cost could matter more than raw benchmark scores.

The launch also signals how quickly Google is iterating. Four Gemini 4 models have now arrived within about seven weeks, and the model picker in the Gemini app has been reorganized around them. For AI Pro and Ultra subscribers, the practical change is immediate: the model they reach for for deeper analysis is now Neon, not Argon or 3.1 Pro.

Sources

Similar Posts

Leave a Reply

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