NewsStocksGoogle DeepMind Introduces Gemini 4 Neon: Frontier-Level Intelligence With Faster, Lower-Cost Task Completion

Google DeepMind Introduces Gemini 4 Neon: Frontier-Level Intelligence With Faster, Lower-Cost Task Completion

Author: Google DeepMind Blog·

Key Takeaways

  • •Google DeepMind launched Gemini 4 Neon on November 18, 2026, positioning it between Carbon for the hardest long-horizon projects and Hydrogen for fast everyday execution within the Gemini 4 family.
  • •The model ranks third on the Artificial Analysis Intelligence Index with a score of 61 and second on the Vals Index, where it sits on the Pareto frontier, meaning no competing model pairs a higher score with a lower cost per task.
  • •Neon is priced at $3 per million input tokens, $0.15 per million cached input tokens, and $15 per million output tokens, with a 1 million-token context window and four thinking levels.
  • •On cybersecurity benchmarks, Neon for cyber achieved state-of-the-art results, including 90.5% on SEC-bench Pro and 51.26% on WebExploitBench v1.2, with access for trusted defenders rolling out through the Fairwind Program.
  • •DeepMind reports improved safety and alignment metrics, including 0.0% cheating on Terminal-Bench 4.0, the lowest fabrication rate recorded in internal Antigravity testing, and a 0.0% attack success rate against indirect prompt injection on Gray Swan's benchmark at the highest thinking level.
Google DeepMind Introduces Gemini 4 Neon: Frontier-Level Intelligence With Faster, Lower-Cost Task Completion

Google DeepMind introduced Gemini 4 Neon on November 18, 2026, a new model the company says pairs frontier-level intelligence with fast end-to-end task completion across complex software engineering, knowledge work, and science workflows. The announcement was authored by Koray Kavukcuoglu, SVP of Google DeepMind and Chief AI Architect at Google. Two threads run through the announcement — where Neon lands on independent leaderboards and what each completed task costs — and they are the same levers teams weigh when deciding whether a model can realistically carry production agent workloads.

Benchmark results, pricing, and availability

According to Google DeepMind, Neon ranks ahead of comparable models and third overall with a score of 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level, landing just behind larger, slower models. On the Vals Index, Neon ranks second and lands on the Pareto frontier — a position meaning no competing model pairs a higher score with a lower cost per task. Beyond raw capability, the company says it is continuing to make its models better to work with: Neon ranks second on steerability — adapting cleanly to mid-task user corrections — on the new Agent Arena Index. The model also excels at multi-step instruction following and builds on DeepMind's ongoing progress to curb fabrication, recording the company's lowest fabrication rate yet in internal Google Antigravity testing.

Neon is priced at $3 per million input tokens, with cached input tokens available at $0.15 per million (a 95% discount) and output tokens at $15 per million. It offers a 1 million-token context window. The cached-input discount maps directly onto the agent workflows described later in the announcement, in which long-running sessions repeatedly re-read large context, while the wide window is what makes the single-context synthesis of spreadsheets, PDFs, charts, audio, and video possible.

The model is rolling out today in the Gemini API in Google AI Studio, in Gemini for business (via the API and model dropdown), for paid Google AI Pro and Ultra subscribers in Google Antigravity, and for Pro and Ultra subscribers in the Gemini app.

Delivers across enterprise, analytical, and scientific work

Enterprise work lives in spreadsheets, PDFs, charts, audio, and video. DeepMind says Neon can synthesize across all of these inputs in a single 1 million-token context window and turn them into structured deliverables. Independent evaluations rank Neon among the top models: it places second on the Vals Index with a score of 70.9%, and first on both the Legal Research Bench and the Tax Agent Bench.

Neon also shows strong wins across categories on Surge AI, ranking first on most benchmarks covering finance, accounting, spreadsheet modeling, and complex constraint following. It scores 73.5% on Chartography for reading dense professional charts and performs on par with leading models on Agents' Last Exam v1.1 (75.8%). It achieves first place overall on both Zapier's AutomationBench at 56.3% and Andon Labs' Vending-Bench 2 at $16,843 — the latter a long-horizon simulation of running a business. It also debuts second globally with a score of 1336 on Arena's new Agent Arena Index, based on more than 4,200 real-world agent sessions.

On scientific and mathematical reasoning — spanning quantitative analysis, mathematics, and technical research problems — Neon is best in class on Terminal-Bench Science (70.5%), RiemannBench (84.0%), and HLE-Diamond (71.3%).

Built for production-ready code

Engineered for long-running agent workflows, Neon sustains performance through complex, tool-heavy tasks end to end, whether running standalone or acting as a fast, high-judgment Orchestrator directing parallel Hydrogen subagents in Google Antigravity. It cuts down on redundant file reads, shares progress as it works, and handles context compaction reliably in long sessions.

Across core developer and researcher workflows, Neon achieves frontier performance and is competitive with the leading models in its class. It ranks first overall on the Artificial Analysis Coding Agent Index with a score of 58.6, posts 65.2% on Terminal-Bench 4.1, and scores 46.2% at its default high setting on FrontierCode 2.0. For production-ready code, Neon handles multi-file debugging and refactoring across repositories with targeted, surgical edits. On SWE-sweep, it already leads on the mini-SWE harness at 12.8% and more than triples that result to a state-of-the-art 39.0% in Google Antigravity, which DeepMind says shows what the Antigravity harness unlocks.

DeepMind also highlighted demonstration projects built with Neon in Google Antigravity:

  • Driving Blender MCP, Gemini 4 Neon used /swarm to build a photorealistic 3D lab bench visualizing model names with elemental accuracy, pairing Bunsen flames for Gemini 4 Carbon and Gemini 4 Hydrogen with high-voltage gas discharge tubes for Gemini 4 Argon and Neon.
  • The model wrote a first-person WebGL2 freeride ski simulator from scratch on real Mont Blanc and Chamonix elevation terrain, combining 10,200 lines of custom shaders, 120 Hz ski physics across six alpine lines, dynamic carved tracks, slab avalanches, and spatial Web Audio with zero pre-made assets.
  • Using /swarm, Neon pulled archival photos to recreate Alan Turing's 1936 study at King's College, Cambridge in interactive 3D — from King's Chapel outside to Room X.8 inside — complete with a working Universal Turing Machine running on the desk.
  • A Gemini 4 Neon agent swarm turned a single image into a 100,000-cell Google Sheet with zero embedded images and zero Apps Script, using more than 500,000 live spreadsheet formulas to run edge detection, map 3D depth, and relight the scene across six interactive modes.

Efficiency and control for production-grade workflows

DeepMind says it is refining thinking levels with each new model release to give teams finer control and a smoother progression from one tier to the next. Neon offers four thinking levels that scale steadily, letting teams dial intelligence and token spend to what each workload needs. That control reflects how production teams actually consume models: per-task cost, not the per-token sticker price, is the figure that scales with usage.

Across third-party benchmarks, Neon delivers high performance at a lower cost per task. At its default high thinking level, it pairs a high Artificial Analysis Intelligence Index score with a 32% lower cost per task than Carbon. It also sits on the Pareto frontier on both the Vals Index and Agent Arena at high, with roughly 50% to 80% lower cost per task than top competing frontier models. The xhigh thinking level adds deeper, multi-step reasoning for the hardest engineering, analytical, and scientific workloads: Neon posts its highest intelligence score at this level while still costing 36% or more less per task than competing models at their highest settings.

Built for fast, efficient, frontier-level cyber defense

As agentic coding scales, DeepMind notes that generating secure code from the start is critical. Gemini 4 Neon for cyber produces more secure code than Carbon across all easy, medium, and hard challenges on AutoBaxBench, achieving state of the art on the hard tier at 76%.

Securing systems also means proactively finding flaws before attackers do, and penetration testing has served as a cornerstone of enterprise cybersecurity. On WebExploitBench v1.2, Neon is state of the art at 51.26%, at half the cost of competing large models.

To fortify existing code, Neon for cyber delivers frontier-level performance across the code hardening lifecycle — vulnerability finding, validating, and patching — while being faster and more cost efficient than large frontier models:

  • Finding code vulnerabilities: Neon achieves state-of-the-art performance on SEC-bench Pro at 90.5%, while costing roughly half as much per challenge.
  • Validating risk through exploitation: to prioritize which vulnerabilities to remediate, defenders rely on models to check which vulnerabilities are exploitable. On ExploitGym, Neon achieves frontier-level performance using half the tokens of Carbon.
  • Patching vulnerabilities: fixing the root cause of vulnerabilities is a critical step in the cybersecurity defense lifecycle. Neon reaches 70.7% on Collinear's CWE-bench v2, driving a 3.4 percentage-point uplift over Carbon.

Gemini 4 Neon for cyber is rolling out to trusted defenders in the coming days as part of the Fairwind Program.

Designed for safe, dependable execution

Gemini 4 Neon builds on improvements from the Gemini 4 safety and security stack, maintaining Gemini 4 Carbon-level protections against severe misuse and prompt injection while advancing further on model alignment and agentic behavior. For a model meant to carry demanding, long-running work in production, DeepMind concentrated on three areas:

Mitigating frontier misuse safety risks. Because Neon brings frontier-level capabilities to a faster, lower-cost model, DeepMind paired it with strong real-time safeguards against cyber and biosecurity misuse comparable to its Carbon protections, while declining far less legitimate work in these areas — refusing, for instance, just 1.98% of tasks on Terminal-Bench 4.0 and 0% on Terminal-Bench-Science.

Advancing alignment and disciplined model behavior. In internal coding use, alert rates from DeepMind's chain-of-thought and action misalignment monitor remained below the already low baselines set by Carbon, suggesting improved alignment propensity for Neon, and that same monitor is deployed to protect Neon across external surfaces. Neon is trained to complete multi-step work without cutting corners, fabricating results, or overstepping user boundaries, continuing to improve the balance between following through on requests and staying within scope. It matched Carbon with 0.0% cheating on Terminal-Bench 4.0 and took shortcuts on just 2.45% of tasks on DeepMind's coding cheating evaluation — the lowest rate of any Gemini 4 model. It also achieved a 0.09% misaligned action rate on DeepMind's agentic safety suite and a 99.2% pass rate on its agent honesty evaluation. On the agentic behavior evaluation, Neon builds on Carbon's strong baseline by further reducing under-service by 43%–56% and overreach by 15%–35% across different effort levels.

Protecting users from indirect prompt injection attacks. On Gray Swan's IPI benchmark, Neon is more resilient than other models in its class when malicious instructions are planted in the web pages, emails, documents, or code it accesses. With real-time safeguards at its highest thinking level, Neon resisted every attack within the Gray Swan benchmark, matching the 0.0% attack success rate reached only by other Gemini models.

Further details are available in the Gemini 4 Neon model card.

Built for deeper analysis in the Gemini app

In the Gemini app, Gemini 4 Neon is aimed at work that requires frontier-level reasoning and fast responses:

  • Higher-quality writing and knowledge work: Neon writes clearly, with well-structured responses and a natural, direct tone. It reads uploads such as spreadsheets, PDFs, and photos more precisely, turns dense material into helpful reports and summaries, and pulls in relevant details from Google apps like Gmail and Docs.
  • Reliable research and analysis: Neon frequently incorporates details from sources across the web to keep responses current on fast-moving topics, such as comparing pricing across vendors before a major purchase or pulling together a complex brief with the latest market data.
  • Faster help for tougher tasks: tasks in Gemini can now run on Neon, working through complex, multi-step jobs more quickly and efficiently than with larger models. Users can get more done while stretching usage limits further when handing off tedious to-dos, such as analyzing billing statements to spot why a utility charge spiked or cross-checking multiple receipts and documents to negotiate a refund.

Where Neon sits in the Gemini 4 family

Neon is built for complex coding, analytical reasoning, and production-grade enterprise workflows. Within the Gemini 4 family, Carbon tackles the hardest, long-horizon projects and Hydrogen handles fast everyday execution, while Neon balances depth and efficiency — delivering frontier-level reasoning for complex, high-frequency workflows faster and more cost efficiently. For teams, the tiering turns into a routing decision: matching each workload's difficulty and frequency to the model whose per-task economics fit it best.

Rollout details

Gemini 4 Neon begins rolling out today:

  • For developers: build with the Gemini API paid tier in Google AI Studio and Android Studio; paid Google AI Pro and Ultra subscribers can try Neon in Google Antigravity, and all Pro and Ultra subscribers can try it in AI Studio.
  • For enterprises: use Gemini 4 Neon in Gemini for business, via the API or model drop-down, and in Google Antigravity.
  • For consumers: Google AI Pro and Ultra subscribers can select Gemini 4 Neon from the model drop-down in the Gemini app.
  • For cybersecurity defenders: eligible Fairwind Program members will begin receiving access over the next few days; to request access to the Fairwind Program, contact your Cloud account team.

About the author

Koray Kavukcuoglu is one of the world's foremost experts in artificial intelligence. As SVP of Google DeepMind, he leads the development of Google DeepMind's state-of-the-art generative AI models and their integration into Google's products at scale. Prior to his current role, he was VP of Research at Google DeepMind, where he founded the deep learning team and his team pioneered major research breakthroughs such as DQN (Deep Q-Network) and WaveNet.

The original announcement is available on the Google DeepMind blog at https://deepmind.google/blog/introducing-gemini-4-neon/