Google introduces Gemini 4 Neon, priced at $3 per million input tokens, rolling out in the Gemini API and app
Google blog post, 18 November 2026: Gemini 4 Neon is presented as a frontier-level model for software engineering, knowledge work and science, with a 1 million-token context window. Benchmark rankings are from third-party indices as cited by Google, plus Google's internal testing; they are not independent verification.
Google announced Gemini 4 Neon on 18 November 2026 in a blog post by Koray Kavukcuoglu, SVP of Google DeepMind and Chief AI Architect at Google. The post presents Neon as a fast frontier-level model for complex software engineering, knowledge work and science workflows, and gives concrete pricing and availability details. The capability claims and rankings below are as stated by Google; where the company cites third-party indices, the post names them.
Stated rankings and testing
According to Google's post, Neon ranks #3 overall with a score of 61 on the Artificial Analysis Intelligence Index at its "xhigh" thinking level, behind what Google describes as larger, slower models; ranks #2 (70.9%) on the Vals Index; ranks #2 on steerability on the new Agent Arena Index; and reaches #1 on the Legal Research Bench and Tax Agent Bench, plus #1 on most Surge AI benchmarks across finance, accounting and spreadsheet modeling. Google also says Neon recorded its lowest fabrication rate yet in internal Google Antigravity testing; that figure is internal and not independently published.
Pricing and availability
The post prices Neon at $3 per million input tokens, $0.15 per million cached input tokens (described as a 95% discount) and $15 per million output tokens, with a 1 million-token context window. Google says Neon is rolling out from 18 November in the Gemini API in Google AI Studio, in Gemini for business (via 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.
What is not yet known
The post does not publish a technical report, parameter count, training details or safety evaluation results beyond the internal fabrication claim. Third-party index positions are as cited by Google on launch day and can change as indices update; we did not independently re-run the benchmarks. Real-world performance against the specific claims, and any regional or tier restrictions beyond those stated, are not addressed in the post.