Google DeepMind introduced Gemini 4 Neon on November 18, 2026, positioning the model as frontier-level intelligence built for fast, end-to-end task completion across software engineering, knowledge work and scientific workflows. Google says Neon ranks third overall on the Artificial Analysis Intelligence Index with a score of 61 at its highest thinking level, behind larger and slower models, and ranks second on the Vals Index, where it sits on the Pareto frontier. The company also highlighted steerability, measured on a new Agent Arena Index, and what it describes as its lowest fabrication rate yet in internal testing within Google Antigravity.
For enterprise and analytical work, Google points to a 1 million-token context window that can combine spreadsheets, PDFs, charts, audio and video in a single session. The company reports first place on the Legal Research Bench and Tax Agent Bench, first overall on Zapier's AutomationBench and Andon Labs' Vending-Bench 2, and a 65.2 percent score on Terminal-Bench 4.1 for coding. Neon is priced at $3 per million input tokens, $15 per million output tokens and $0.15 per million cached input tokens. It is rolling out through the Gemini API, Google AI Studio, Gemini for business, Google Antigravity for paid AI Pro and Ultra subscribers, and the Gemini app for Pro and Ultra subscribers.
Many of the headline results come from leaderboards run by third parties, but others, such as the 39.0 percent score on SWE-sweep inside Google's own Antigravity harness, depend on Google's tooling. The benchmark comparisons are therefore best treated as vendor-reported until teams test the model on their own workloads. For Australian businesses and professionals, the release changes little immediately unless they already build on Gemini APIs or use Gemini for business. The token pricing is competitive, and Google says Neon at its default setting costs 32 percent less per task than its Carbon model, though that is Google's own comparison. Any organisation sending personal information to an overseas AI API should check the provider's data handling terms against the Privacy Act 1988, particularly Australian Privacy Principle 8 on overseas disclosure. A low-risk test on an existing spreadsheet or document workflow, compared against the current model, is the practical next step.
Key Points
Google says Gemini 4 Neon ranks third overall on the Artificial Analysis Intelligence Index and second on the Vals Index
The model offers a 1 million-token context window and is priced at $3 per million input tokens and $15 per million output tokens
Google reports first place on several enterprise and agent benchmarks, including Zapier AutomationBench and Vending-Bench 2
Neon is rolling out in the Gemini API, Google AI Studio, Gemini for business, Google Antigravity and the Gemini app for paid tiers
Most performance claims are vendor-reported, so independent testing on real workloads matters before adoption