Gemini 4 Neon: Carbon-Level Brains, Flash-Like Speed

Human Verified

Share to: , and

A glowing green and cyan neural network sphere above a dark desk

Google introduced Gemini 4 Neon on November 18, a mid-tier model that pairs frontier-level reasoning with faster, cheaper task completion across coding, knowledge work and science. It costs 3 dollars per million input tokens and 15 dollars per million output tokens.

Where Neon Sits

Neon sits between Carbon, built for the hardest long-horizon projects, and Hydrogen, built for fast everyday work. It ranks third overall with a score of 61 on the Artificial Analysis Intelligence Index at its highest thinking level, just behind larger, slower models.

The model also ranks second on the Vals Index and first on the Artificial Analysis Coding Agent Index, with a score of 58.6. It carries a one-million-token context window and four thinking levels, letting teams dial intelligence and token spend to each workload.

Built for Code and Science

On coding, Neon scores 65.2 percent on Terminal-Bench 4.1 and 46.2 percent on FrontierCode 2.0 at its default high setting. In Google Antigravity, where it can direct parallel Hydrogen subagents as an orchestrator, it reaches a state-of-the-art 39 percent on SWE-sweep.

In science and analysis, Neon is best in class on Terminal-Bench Science at 70.5 percent, RiemannBench at 84 percent and HLE-Diamond at 71.3 percent. It also ranks first on the Legal Research and Tax Agent benches, and scores 73.5 percent on Chartography.

Cheaper Than Carbon

At its default high thinking level, Neon costs 32 percent less per task than Carbon, Google said, and roughly 50 to 80 percent less than top competing frontier models. At its highest setting it still costs 36 percent less per task than rivals.

Neon is rolling out now in the Gemini API in Google AI Studio, in Gemini for business, in Antigravity and in the Gemini app for paid Pro and Ultra subscribers. It replaces Gemini 3.1 Pro, and Tasks in Gemini can now run on Neon too.

Carbon and Argon cost 2 dollars per million input tokens and 10 dollars per million output tokens until November 30, under promotional pricing. From December 1 they rise to 4 dollars and 20 dollars, making Neon 25 percent cheaper than both on a per-token basis.

Cyber and Safety

A cyber-focused version, Neon for cyber, produces more secure code than Carbon across easy, medium and hard challenges on AutoBaxBench, reaching state of the art on the hard tier at 76 percent. It reaches trusted defenders through Google’s Fairwind Program in the coming days.

Neon builds on Carbon-level protections against severe misuse and prompt injection, while recording Google’s lowest fabrication rate yet in internal Antigravity testing. On Gray Swan’s prompt-injection benchmark it resisted every attack at its highest thinking level, matching a zero percent attack success rate.

Our Read

Neon is Google slicing the Gemini 4 family by price and performance, and it is the tier most developers will actually use. Carbon stays the flagship for the hardest jobs, and Neon takes the middle ground where most production work lives.

The benchmark story is strong but mostly Google’s own, with independent Artificial Analysis and Vals scores as the check. The real test is whether developers move work from Carbon to Neon once the pricing ends on December 1, and whether Claude and GPT-6.2 answer with cheaper tiers.

Why It Matters

A frontier-capable model at half the cost of the top tier changes the economics of agentic coding and research. If Neon holds its scores in real use, it becomes the default for production workloads, and rivals will have to answer on price.

Background

Carbon and Hydrogen launched the week before Neon, and Google rolled Hydrogen into AI Mode on November 12 as its biggest search upgrade. Neon completes the family for now, with Carbon still the flagship and Hydrogen the fast everyday engine.

Source: Google’s announcement

Share to: , and

Faheem R. Khattak

Senior automation engineer and AI marketing journo covering LLMs, algorithms, analytics, core updates, Bing search, and technical SEO.

0 Comments

Leave a Reply

Optional Take Tag:

🛡️ Spam protection: Honeypot + Time-Trap Active

This site uses Akismet to reduce spam. Learn how your comment data is processed.