Thursday, November 19, 2026 Updated Briefings Saved 中文
AI Daily Radar

AI and the frontier tech around it, explained every morning

Released

Google releases Gemini 4 Neon with 1M-token context, higher pricing

Photo: Wesley Nitsckie / Wikimedia Commons · CC BY-SA 2.0, cropped

Google released Gemini 4 Neon, a frontier model for software engineering, knowledge work and science, and is rolling it out in the API, the Antigravity coding platform and the Gemini app. According to The Decoder, Neon handles text, images, video, audio and PDFs with a one-million-token context window and costs $3 per million input tokens and $15 per million output tokens, above earlier Gemini 4 models. The Decoder described it as Google’s fourth new Gemini model in a month, while Google’s announcement emphasizes “fast end-to-end task completion” rather than listing benchmark results in the available reporting.

Verified and unconfirmed

  • Verified: Google’s blog says Gemini 4 Neon is rolling out in the API, Antigravity and the Gemini app for coding, knowledge work and science, with “fast end-to-end task completion.”
  • Not yet confirmed: The one-million-token context window, $3/$15 pricing and “fourth new Gemini model in a month” framing come from The Decoder, not the quoted Google post; the available reporting includes no independent benchmark results, model card or dated general-availability schedule.

Background

A context window is the amount of text, images or other input a model can consider in a single request; a one-million-token window is meant to let a model work across large codebases or document sets without breaking them into chunks. Google has shipped several Gemini 4 variants in quick succession, and the Hacker News discussion refers to earlier Carbon, Hydrogen and Argon models, so users now have to choose among siblings rather than a single flagship.

Why it matters

For teams using Gemini through the API, the immediate change is cost: at $3 per million input tokens and $15 per million output tokens, Neon sits above earlier Gemini 4 models, so production workloads need a fresh cost check before switching. Developers can test it in Antigravity and the Gemini app, but the Hacker News thread suggests comparing it with Carbon on long-horizon research tasks rather than treating the newest model as a strict upgrade.

Community discussion

Hacker News commenters focused on naming fatigue and lineup positioning: one said Google, Anthropic and OpenAI have created a confusing map of model names, while another argued Carbon remains the “long horizon” model and scores higher on research-scale evals even as Neon leads on the AA Index overall. A separate commenter reported that Neon, Carbon and Hydrogen used three to four times more tokens than Argon for the same demo-building task — a single user’s experience, not a benchmark.

Coverage of this story

Comments

Comments

Comments use a GitHub account and are moderated. Comment rules

More in AI

See all