Google launches Gemini 4 Neon at $3 per million input tokens
Google released Gemini 4 Neon on Nov 18, 2026, priced at $3 per million input tokens.

· Last verified: 19 Nov 2026 (Google (The Keyword))

Summary
- Google released Gemini 4 Neon on Nov 18, 2026, priced at $3 per million input tokens.
- Google says Neon costs 32% less per task than Carbon at its default high thinking level.
- Cheaper frontier-level coding and document work lowers the bill for running production AI agents.
The latest
Gemini 4 Neon went live Tuesday at $3 per million input tokens and $15 per million output tokens, with a 1 million-token context window. Google says the model delivers frontier-level intelligence with fast end-to-end task completion across software engineering, knowledge work and science workflows. Koray Kavukcuoglu, SVP at Google DeepMind and the company's chief AI architect, announced the release.
Details
- The pricing: Input runs at $3 per million tokens and output at $15 per million. Cached input tokens cost $0.15 per million, which Google describes as a 95% discount. The 1 million-token context window takes spreadsheets, PDFs, charts, audio and video in a single pass.
- The cost claim: At its default high thinking level, Google says Neon pairs a high Artificial Analysis Intelligence Index score with a 32% lower cost per task than Carbon. On the Vals index and Agent Arena at high, Google puts the gap at roughly 50% to 80% below top competing frontier models.
- Benchmark rankings: Google says Neon scores 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level, third overall behind larger and slower models. It ranks second on the Vals Index at 70.9%, and first on both the Legal Research Bench and Tax Agent Bench.
- Coding results: Google reports Neon first overall on the Artificial Analysis Coding Agent Index at 58.6, with 65.2% on Terminal-Bench 4.1 and 46.2% on FrontierCode 2.0 at its default high setting. On SWE-sweep it leads the mini-SWE harness at 12.8%, rising to 39.0% inside Google Antigravity.
- Agent workloads: Google says Neon took first on Zapier's AutomationBench at 56.3% and on Andon Labs' Vending-Bench 2 with $16,843 in a long-horizon business simulation. It debuted second globally at 1,336 on Arena's new Agent Arena Index, drawn from more than 4,200 real-world agent sessions.
- Science scores: On scientific and mathematical reasoning, Google places Neon best in class on Terminal-Bench Science at 70.5%, RiemannBench at 84.0% and HLE-Diamond at 71.3%. It reports 73.5% on Chartography for dense professional charts and 75.8% on Agents' Last Exam v1.1.
- Security version: A separate build, Gemini 4 Neon for cyber, reaches 76% on the hard tier of AutoBaxBench and 51.26% on WebExploitBench v1.2, according to Google. The company says it hits 90.5% on SEC-bench Pro at roughly half the cost per challenge and 70.7% on Collinear's CWE-bench v2.
- Safety figures: Google reports 0.0% cheating on Terminal-Bench 4.0, shortcuts on 2.45% of tasks in its coding cheating evaluation, a 0.09% misaligned action rate on its agentic safety suite and a 99.2% pass rate on agent honesty. It says Neon refused 1.98% of Terminal-Bench 4.0 tasks.
- Where it ships: Neon is available in the Gemini API through Google AI Studio and Android Studio, in Gemini for business, in Google Antigravity for paid AI Pro and Ultra subscribers, and in the Gemini app for Pro and Ultra users. The cyber build reaches Fairwind Program members over the coming days.
Background
Neon slots between two existing Gemini 4 models: Carbon, which Google assigns to its hardest long-horizon projects, and Hydrogen, built for fast everyday execution. Google positions Neon as the balance of depth and efficiency for high-frequency enterprise workflows.
Between the lines
The headline number is price, not capability. Neon ranks third on Google's own intelligence index, behind models the company concedes are larger and slower, but the pitch rests on cost per task — 32% under Carbon, and half the token spend on some security evaluations. For teams running agents continuously, the recurring bill matters more than a benchmark position. Nearly every figure cited comes from Google's own testing.
What's next
Watch for independent benchmark runs outside Google's own evaluations, and for Fairwind Program access to the cyber build over the coming days. Competing frontier labs' response on price per token is the next signal.
Source: Google (The Keyword)