Google is not slowing down. Just six days after Gemini 4 Hydrogen arrived, Google DeepMind has introduced Gemini 4 Neon, the fourth model in the Gemini 4 family — and the one aimed squarely at people who actually use AI all day. Announced on November 18, 2026 by Koray Kavukcuoglu, SVP of Google DeepMind and Google’s Chief AI Architect, Neon promises “frontier-level intelligence with fast end-to-end task completion” across software engineering, knowledge work and science. The headline number for most users is simpler: it is rolling out now to Google AI Pro subscribers, not just the top-tier Ultra plan, at $3 per million input tokens.
What Is Gemini 4 Neon?
Neon sits in the middle of Google’s fast-growing Gemini 4 lineup. Carbon, released November 10, is the heavyweight for the hardest problems. Hydrogen, which launched November 12, is the fast, low-cost model for everyday agentic work. Neon is the balance between the two: strong enough for complex, long-running work, but built for speed and cost rather than maximum capability. Like its siblings, it carries a 1 million-token context window, so it can hold spreadsheets, PDFs, charts, audio and video in a single request and turn them into structured deliverables. It accepts text, image, video, audio and PDF inputs and returns text.
Gemini 4 Neon Pricing: $3 In, $15 Out

The pricing is where Neon gets interesting — and a little strange. Developers pay $3 per million input tokens and $15 per million output tokens, with cached input at just $0.15 per million, a 95% discount. That is exactly double Hydrogen’s $1.50/$7.50, and it only makes sense on a calendar: Carbon’s introductory price of $2 input and $10 output runs through November 30, making Carbon temporarily cheaper per token than Neon. When that introductory period ends on November 30, Carbon rises to $4 input and $20 output, and Neon becomes the cheaper option per token as well. Google’s real pitch, though, is cost per task, not cost per token: at its default “high” thinking level, Google says Neon costs 32% less per task than Carbon, and 50% to 80% less per task than top competing frontier models. Independent evaluator Artificial Analysis puts Neon’s cost at $6.44 per task — about 17% less than Carbon’s $7.76 — and says it finishes tasks in about 21 minutes, roughly 40% faster than Carbon’s 36. For a sense of how fast inference economics are moving across the industry, see also Cerebras’ disaggregated inference approach, where GPUs handle prompt processing and wafer-scale chips generate the answer.
Benchmarks: Third Overall, First in Coding Agents
On the benchmarks Google chose to highlight, Neon looks like a genuine middleweight contender. It scored 61 on the Artificial Analysis Intelligence Index at its highest “xhigh” thinking level, ranking third overall — one point ahead of Carbon. The coding numbers are the strongest part of the story: Neon ranks first on the Artificial Analysis Coding Agent Index at 58.6, with 65.2% on Terminal-Bench 4.1, and Google says it recorded its lowest fabrication rate yet in internal testing inside Antigravity, its agentic coding tool. It also claims second place on the Vals Index (70.9%), first place on the Legal Research Bench and Tax Agent Bench, and a score of 1336 on Arena’s new Agent Arena Index, ranking second globally and second on steerability — how well it handles corrections mid-task. Other claimed wins include first place on Zapier’s AutomationBench (56.3%) and Andon Labs’ Vending-Bench 2 business simulation ($16,843), plus best-in-class results on Terminal-Bench Science (70.5%), RiemannBench (84.0%) and HLE-Diamond (71.3%). As with any vendor announcement, these figures come from Google’s own testing and from evaluators Google cites, so treat them as a strong signal rather than independent proof. For comparison, the rival frontier launch this week was OpenAI’s GPT-6.2 Astra.
Built for Agentic Coding: The Orchestrator Model

Neon is designed for long-running agent work, and it has a trick its siblings do not advertise as loudly: it can work alone or act as an “Orchestrator” that directs parallel Hydrogen subagents inside Antigravity. The idea is that Neon plans and supervises while cheaper, faster Hydrogen models do the parallel grunt work. On the SWE-sweep benchmark, Google reports Neon scored 12.8% on the mini-SWE harness and 39.0% inside Antigravity. Four thinking levels let teams trade intelligence against token spend, and at the top “xhigh” setting Google says Neon still costs at least 36% less per task than rivals at their highest settings. If your workflow is multi-step coding, research or document production rather than one-shot questions, that orchestrator role is the feature most likely to matter.
Gemini 4 Neon for Cyber Is Coming Too
Google is also preparing a cybersecurity-tuned version, Gemini 4 Neon for cyber, which will reach trusted defenders in the coming days through the company’s Fairwind Program. Google says it writes more secure code than Carbon on AutoBaxBench, scoring 76% on that benchmark’s hard tier, and reports 90.5% on SEC-bench Pro for finding vulnerabilities and 70.7% on Collinear’s CWE-bench v2 for patching them. On safety, Google says Neon keeps the same protections against severe misuse and prompt injection as Carbon, while turning down far fewer legitimate requests in sensitive areas — refusing 1.98% of tasks on Terminal-Bench 4.0 and none on Terminal-Bench-Science. That combination matters in a week when Google’s own threat researchers have warned that AI agents are already being used to turn patches into exploits, as covered in our report on Google’s warning about autonomous AI attacks. The cyber-defence angle also connects to OpenAI’s Daybreak cyber defence deal with NATO.
How to Get Gemini 4 Neon
Neon is rolling out now in the Gemini API in Google AI Studio and in Gemini for business, through both the API and the model dropdown. Paid Google AI Pro and Ultra subscribers can also use it in Google Antigravity and in the Gemini app, where it appears in the model selector as the rollout reaches each account. Google has not said whether Neon will come to free Gemini users or to AI Mode in Search, where Carbon and Hydrogen are already available. The next date to watch is November 30, when Carbon’s introductory API pricing ends and Neon’s per-token price becomes the cheaper of the two. The Gemini app itself keeps gaining new abilities — Google Gemini Tasks goes free for everyone. For more details, see TalkAndroid’s coverage.
Should You Switch to Gemini 4 Neon?
If you already pay for Google AI Pro, Neon is the most capable model Google has ever put on that plan, and there is no reason not to try it in the model selector as soon as it appears. For developers, the decision is a cost-per-task calculation: if your work is long, multi-step agent runs, Neon’s orchestrator mode and faster completion times may beat Hydrogen’s cheaper tokens. If you only need quick everyday answers, Hydrogen remains the cheaper tool. And if you need maximum capability regardless of cost, Carbon is still the flagship — at least until its price rises on November 30. Either way, Google has now shipped four Gemini 4 models in about seven weeks, and the pace shows no sign of slowing.
The Bottom Line
Gemini 4 Neon is Google’s bet that the middle of the AI market is where the real work happens: frontier-class intelligence, fast enough for daily professional use, priced so teams can run it constantly rather than sparingly. With top coding-agent scores, a 1 million-token context window and a cyber-tuned sibling on the way, it is the most interesting Gemini 4 release yet for working developers — and the first one Pro subscribers get on day one.

